2021 |
Buroni, Giovanni; Bontempi, Gianluca; Determe, Karl A tutorial on network-wide multi-horizon traffic forecasting with deep learning Journal Article CEUR Workshop Proceedings, 2841 , 2021, (Language of publication: en). @article{info:hdl:2013/322517, title = {A tutorial on network-wide multi-horizon traffic forecasting with deep learning}, author = {Giovanni Buroni and Gianluca Bontempi and Karl Determe}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/322517/3/EasyChair-Preprint-4944.pdf}, year = {2021}, date = {2021-01-01}, journal = {CEUR Workshop Proceedings}, volume = {2841}, abstract = {Traffic flow forecasting is fundamental to today's Intelligent Transportation Systems (ITS). It involves the task of learning traffic complex dynamics in order to predict future conditions. This is particularly challenging when it comes to predict the traffic status for multiple horizons into the future and at the same time for the entire transportation network. In this context deep learning models have recently shown promising results. This models can inherently capture the non-linear space-temporal correlations (ST) in traffic by taking advantage of the huge volume of data available. In this study the authors present a LSTM encoder-decoder for multi-horizon traffic flow predictions. We adopted a direct approach in which the model simultaneously predict traffic conditions for the entire Belgian motorway transport network at each time step. The results clearly show the superiority of this model when compared with other deep learning models. In the workshop, conference attendees will learn how to process and visualize mobility data, obtain optimal features for traffic flow forecasting, build a LSTM encoder-decoder and perform predictions in an online manner.}, note = {Language of publication: en}, keywords = {}, pubstate = {published}, tppubtype = {article} } Traffic flow forecasting is fundamental to today's Intelligent Transportation Systems (ITS). It involves the task of learning traffic complex dynamics in order to predict future conditions. This is particularly challenging when it comes to predict the traffic status for multiple horizons into the future and at the same time for the entire transportation network. In this context deep learning models have recently shown promising results. This models can inherently capture the non-linear space-temporal correlations (ST) in traffic by taking advantage of the huge volume of data available. In this study the authors present a LSTM encoder-decoder for multi-horizon traffic flow predictions. We adopted a direct approach in which the model simultaneously predict traffic conditions for the entire Belgian motorway transport network at each time step. The results clearly show the superiority of this model when compared with other deep learning models. In the workshop, conference attendees will learn how to process and visualize mobility data, obtain optimal features for traffic flow forecasting, build a LSTM encoder-decoder and perform predictions in an online manner. |
Bontempi, Gianluca; Chavarriaga, Ricardo; eD Canck, Hans; Girardi, Emanuela; Hoos, Holger H; Kilbane-Dawe, Iarla; Ball, Tonio; Nowe, Ann; Sousa, Jose; Bacciu, Davide; Aldinucci, Marco; eD Domenico, Manlio; Saffiotti, Alessandro; Maratea, Marco The CLAIRE COVID-19 initiative: approach, experiences and recommendations Journal Article Ethics and information technology, 2021, (DOI: 10.1007/s10676-020-09567-7). @article{info:hdl:2013/321123, title = {The CLAIRE COVID-19 initiative: approach, experiences and recommendations}, author = {Gianluca Bontempi and Ricardo Chavarriaga and Hans eD Canck and Emanuela Girardi and Holger H Hoos and Iarla Kilbane-Dawe and Tonio Ball and Ann Nowe and Jose Sousa and Davide Bacciu and Marco Aldinucci and Manlio eD Domenico and Alessandro Saffiotti and Marco Maratea}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/321123/3/Bontempi2021_Article_TheCLAIRECOVID-19InitiativeApp-2.pdf}, year = {2021}, date = {2021-01-01}, journal = {Ethics and information technology}, abstract = {A volunteer effort by Artificial Intelligence (AI) researchers has shown it can deliver significant research outcomes rapidly to help tackle COVID-19. Within two months, CLAIRE's self-organising volunteers delivered the World's first comprehensive curated repository of COVID-19-related datasets useful for drug-repurposing, drafted review papers on the role CT/X-ray scan analysis and robotics could play, and progressed research in other areas. Given the pace required and nature of voluntary efforts, the teams faced a number of challenges. These offer insights in how better to prepare for future volunteer scientific efforts and large scale, data-dependent AI collaborations in general. We offer seven recommendations on how to best leverage such efforts and collaborations in the context of managing future crises.}, note = {DOI: 10.1007/s10676-020-09567-7}, keywords = {}, pubstate = {published}, tppubtype = {article} } A volunteer effort by Artificial Intelligence (AI) researchers has shown it can deliver significant research outcomes rapidly to help tackle COVID-19. Within two months, CLAIRE's self-organising volunteers delivered the World's first comprehensive curated repository of COVID-19-related datasets useful for drug-repurposing, drafted review papers on the role CT/X-ray scan analysis and robotics could play, and progressed research in other areas. Given the pace required and nature of voluntary efforts, the teams faced a number of challenges. These offer insights in how better to prepare for future volunteer scientific efforts and large scale, data-dependent AI collaborations in general. We offer seven recommendations on how to best leverage such efforts and collaborations in the context of managing future crises. |
Bontempi, Gianluca Statistical foundations of machine learning Book 2021, (Language of publication: en). @book{info:hdl:2013/325210, title = {Statistical foundations of machine learning}, author = {Gianluca Bontempi}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/325210/3/syl.pdf}, year = {2021}, date = {2021-01-01}, note = {Language of publication: en}, keywords = {}, pubstate = {published}, tppubtype = {book} } |
Guida, Sibilla Di; Han, The Anh T A H; Kirchsteiger, Georg; Lenaerts, Tom; Zisis, Ioannis Repeated interaction and its impact on cooperation and surplus allocation---an experimental analysis Journal Article Games, 12 (1), 2021, (DOI: 10.3390/g12010025). @article{info:hdl:2013/321049, title = {Repeated interaction and its impact on cooperation and surplus allocation---an experimental analysis}, author = {Sibilla Di Guida and The Anh T A H Han and Georg Kirchsteiger and Tom Lenaerts and Ioannis Zisis}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/321049/1/doi_304693.pdf}, year = {2021}, date = {2021-01-01}, journal = {Games}, volume = {12}, number = {1}, abstract = {This paper investigates how the possibility of affecting group composition combined with the possibility of repeated interaction impacts cooperation within groups and surplus distribution. We developed and tested experimentally a Surplus Allocation Game where cooperation of four agents is needed to produce surplus, but only two have the power to allocate it among the group members. Three matching procedures (corresponding to three separate experimental treatments) were used to test the impact of the variables of interest. A total of 400 subjects participated in our research, which was computer-based and conducted in a laboratory. Our results show that allowing for repeated interaction with the same partners leads to a self-selection of agents into groups with different life spans, whose duration is correlated with the behavior of both distributors and receivers. While behavior at the group level is diverse for surplus allocation and amount of cooperation, aggregate behavior is instead similar when repeated interaction is allowed or not allowed. We developed a behavioral model that captures the dynamics observed in the experimental data and sheds light into the rationales that drive the agents' individual behavior, suggesting that the most generous distributors are those acting for fear of rejection, not for true generosity, while the groups lasting the longest are those composed by this type of distributors and ``undemanding'' receivers.}, note = {DOI: 10.3390/g12010025}, keywords = {}, pubstate = {published}, tppubtype = {article} } This paper investigates how the possibility of affecting group composition combined with the possibility of repeated interaction impacts cooperation within groups and surplus distribution. We developed and tested experimentally a Surplus Allocation Game where cooperation of four agents is needed to produce surplus, but only two have the power to allocate it among the group members. Three matching procedures (corresponding to three separate experimental treatments) were used to test the impact of the variables of interest. A total of 400 subjects participated in our research, which was computer-based and conducted in a laboratory. Our results show that allowing for repeated interaction with the same partners leads to a self-selection of agents into groups with different life spans, whose duration is correlated with the behavior of both distributors and receivers. While behavior at the group level is diverse for surplus allocation and amount of cooperation, aggregate behavior is instead similar when repeated interaction is allowed or not allowed. We developed a behavioral model that captures the dynamics observed in the experimental data and sheds light into the rationales that drive the agents' individual behavior, suggesting that the most generous distributors are those acting for fear of rejection, not for true generosity, while the groups lasting the longest are those composed by this type of distributors and ``undemanding'' receivers. |
`e, Nathaniel Vincent Mon P; Lenaerts, Tom; d dos Pacheco, Jorge Manuel Santos J M S; Dingli, David Multistage feedback-driven compartmental dynamics of hematopoiesis Journal Article iScience, 24 (4), 2021, (DOI: 10.1016/j.isci.2021.102326). @article{info:hdl:2013/322462, title = {Multistage feedback-driven compartmental dynamics of hematopoiesis}, author = {Nathaniel Vincent Mon P{`e}re and Tom Lenaerts and Jorge Manuel Santos J M d S dos Pacheco and David Dingli}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/322462/1/doi_306106.pdf}, year = {2021}, date = {2021-01-01}, journal = {iScience}, volume = {24}, number = {4}, abstract = {Human hematopoiesis is surprisingly resilient to disruptions, providing suitable responses to severe bleeding, long-lasting immune activation, and even bone marrow transplants. Still, many blood disorders exist which push the system past its natural plasticity, resulting in abnormalities in the circulating blood. While proper treatment of such diseases can benefit from understanding the underlying cell dynamics, these are non-trivial to predict due to the hematopoietic system's hierarchical nature and complex feedback networks. To characterize the dynamics following different types of perturbations, we investigate a model representing hematopoiesis as a sequence of compartments covering all maturation stages---from stem to mature cells---where feedback regulates cell production to ongoing necessities. We find that a stable response to perturbations requires the simultaneous adaptation of cell differentiation and self-renewal rates, and show that under conditions of continuous disruption---as found in chronic hemolytic states---compartment cell numbers evolve to novel stable states.}, note = {DOI: 10.1016/j.isci.2021.102326}, keywords = {}, pubstate = {published}, tppubtype = {article} } Human hematopoiesis is surprisingly resilient to disruptions, providing suitable responses to severe bleeding, long-lasting immune activation, and even bone marrow transplants. Still, many blood disorders exist which push the system past its natural plasticity, resulting in abnormalities in the circulating blood. While proper treatment of such diseases can benefit from understanding the underlying cell dynamics, these are non-trivial to predict due to the hematopoietic system's hierarchical nature and complex feedback networks. To characterize the dynamics following different types of perturbations, we investigate a model representing hematopoiesis as a sequence of compartments covering all maturation stages---from stem to mature cells---where feedback regulates cell production to ongoing necessities. We find that a stable response to perturbations requires the simultaneous adaptation of cell differentiation and self-renewal rates, and show that under conditions of continuous disruption---as found in chronic hemolytic states---compartment cell numbers evolve to novel stable states. |
`e, Quentin Rivi; Xiao, Qiying; Gutsch, Annelie; Defrance, Matthieu; Webb, A A R; Verbruggen, Nathalie Mg deficiency interacts with the circadian clock and phytochromes pathways in Arabidopsis Journal Article Annals of Applied Biology, 178 (2), pp. 387-399, 2021, (DOI: 10.1111/aab.12659). @article{info:hdl:2013/322914, title = {Mg deficiency interacts with the circadian clock and phytochromes pathways in Arabidopsis}, author = {Quentin Rivi{`e}re and Qiying Xiao and Annelie Gutsch and Matthieu Defrance and A A R Webb and Nathalie Verbruggen}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/322914/3/MgDeficiencyInPlantsLightClock.pdf}, year = {2021}, date = {2021-01-01}, journal = {Annals of Applied Biology}, volume = {178}, number = {2}, pages = {387-399}, note = {DOI: 10.1111/aab.12659}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
Ciortan, Madalina; Defrance, Matthieu Contrastive self-supervised clustering of scRNA-seq data. Journal Article BMC bioinformatics, 22 (1), pp. 280, 2021, (DOI: 10.1186/s12859-021-04210-8). @article{info:hdl:2013/325214, title = {Contrastive self-supervised clustering of scRNA-seq data.}, author = {Madalina Ciortan and Matthieu Defrance}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/325214/1/doi_308858.pdf}, year = {2021}, date = {2021-01-01}, journal = {BMC bioinformatics}, volume = {22}, number = {1}, pages = {280}, abstract = {Single-cell RNA sequencing (scRNA-seq) has emerged has a main strategy to study transcriptional activity at the cellular level. Clustering analysis is routinely performed on scRNA-seq data to explore, recognize or discover underlying cell identities. The high dimensionality of scRNA-seq data and its significant sparsity accentuated by frequent dropout events, introducing false zero count observations, make the clustering analysis computationally challenging. Even though multiple scRNA-seq clustering techniques have been proposed, there is no consensus on the best performing approach. On a parallel research track, self-supervised contrastive learning recently achieved state-of-the-art results on images clustering and, subsequently, image classification.}, note = {DOI: 10.1186/s12859-021-04210-8}, keywords = {}, pubstate = {published}, tppubtype = {article} } Single-cell RNA sequencing (scRNA-seq) has emerged has a main strategy to study transcriptional activity at the cellular level. Clustering analysis is routinely performed on scRNA-seq data to explore, recognize or discover underlying cell identities. The high dimensionality of scRNA-seq data and its significant sparsity accentuated by frequent dropout events, introducing false zero count observations, make the clustering analysis computationally challenging. Even though multiple scRNA-seq clustering techniques have been proposed, there is no consensus on the best performing approach. On a parallel research track, self-supervised contrastive learning recently achieved state-of-the-art results on images clustering and, subsequently, image classification. |
2020 |
Pollaris, Arnaud; Bontempi, Gianluca Latent Causation: An algorithm for pairs of correlated latent variables in Linear Non-Gaussian Structural Equation Modeling Miscellaneous 2020, (Conference: BNAIC/BENELEARN (19 & 20 November 2020: Leiden (online))). @misc{info:hdl:2013/314680, title = {Latent Causation: An algorithm for pairs of correlated latent variables in Linear Non-Gaussian Structural Equation Modeling}, author = {Arnaud Pollaris and Gianluca Bontempi}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/314680/3/BNAICBENELEARN_2020_Final_paper.pdf}, year = {2020}, date = {2020-01-01}, abstract = {This paper addresses the problem of inferring causation in a pair of linearly correlated continuous latent variables. We first discuss the limitations of the Direction Dependance Analysis (DDA) approach and then introduce the Latent Causation (LC). Five variants (in terms of dependency statistic) of the LC algorithm are assessed with ROC curves, then we consider the case of a latent confounder (uniform or chi-square distributed). While the distribution and the correlations of the latent confounder influence the accuracy, experimental results show the robustness of the method using bootstrapped p-values. Implications and limits of the experimental results are then discussed together with future directions.}, note = {Conference: BNAIC/BENELEARN (19 & 20 November 2020: Leiden (online))}, keywords = {}, pubstate = {published}, tppubtype = {misc} } This paper addresses the problem of inferring causation in a pair of linearly correlated continuous latent variables. We first discuss the limitations of the Direction Dependance Analysis (DDA) approach and then introduce the Latent Causation (LC). Five variants (in terms of dependency statistic) of the LC algorithm are assessed with ROC curves, then we consider the case of a latent confounder (uniform or chi-square distributed). While the distribution and the correlations of the latent confounder influence the accuracy, experimental results show the robustness of the method using bootstrapped p-values. Implications and limits of the experimental results are then discussed together with future directions. |
Simar, Cédric; Petieau, Mathieu; Cebolla, Ana Maria; Leroy, Axelle; Bontempi, Gianluca; Chéron, Guy EEG-based brain-computer interface for alpha speed control of a small robot using the MUSE headband Inproceedings 2020, (DOI: 10.1109/IJCNN48605.2020.9207486). @inproceedings{info:hdl:2013/315071, title = {EEG-based brain-computer interface for alpha speed control of a small robot using the MUSE headband}, author = {Cédric Simar and Mathieu Petieau and Ana Maria Cebolla and Axelle Leroy and Gianluca Bontempi and Guy Chéron}, url = {http://hdl.handle.net/2013/ULB-DIPOT:oai:dipot.ulb.ac.be:2013/315071}, year = {2020}, date = {2020-01-01}, abstract = {Non-invasive BMI applications are increasingly used in different contexts ranging from industrial, clinical and gaming. After having tested the difference between a classical EEG recorder with electroconductive gel (ANT system) and the MUSE EEG headband, we studied the BCI performances of the later during the control of a small robot. We demonstrated that the participants were able to successfully control the robot using an online brain-computer interface based on the signal power in different frequency bands (delta, theta and alpha) characterizing the eyes-opened and relaxed eyes-closed states. Additionally, we performed a correlation analysis which demonstrated that the BCI commands were more related to a delta or theta power decrease for the determination of the classifier output probability and to the alpha power increase for the speed control of the robot.}, note = {DOI: 10.1109/IJCNN48605.2020.9207486}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Non-invasive BMI applications are increasingly used in different contexts ranging from industrial, clinical and gaming. After having tested the difference between a classical EEG recorder with electroconductive gel (ANT system) and the MUSE EEG headband, we studied the BCI performances of the later during the control of a small robot. We demonstrated that the participants were able to successfully control the robot using an online brain-computer interface based on the signal power in different frequency bands (delta, theta and alpha) characterizing the eyes-opened and relaxed eyes-closed states. Additionally, we performed a correlation analysis which demonstrated that the BCI commands were more related to a delta or theta power decrease for the determination of the classifier output probability and to the alpha power increase for the speed control of the robot. |
Caro, Fabrizio De; Stefani, Jacopo De; Bontempi, Gianluca; Vaccaro, Alfredo A; Villacci, Domenico D Robust Assessment of Short-Term Wind Power Forecasting Models on Multiple Time Horizons Journal Article Technology and Economics of Smart Grids and Sustainable Energy, 5 (1), 2020, (DOI: 10.1007/s40866-020-00090-8). @article{info:hdl:2013/314435, title = {Robust Assessment of Short-Term Wind Power Forecasting Models on Multiple Time Horizons}, author = {Fabrizio De Caro and Jacopo De Stefani and Gianluca Bontempi and Alfredo A Vaccaro and Domenico D Villacci}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/314435/1/doi_298079.pdf}, year = {2020}, date = {2020-01-01}, journal = {Technology and Economics of Smart Grids and Sustainable Energy}, volume = {5}, number = {1}, abstract = {The massive penetration of renewable power generation in modern power grids is an effective way to reduce the impact of energy production on global warming. Unfortunately, the wind power generation may affect the regular operation of electrical systems, due to the stochastic and intermittent nature of the wind. For this reason, reducing the uncertainty about the wind evolution, e.g. by using short-term wind power forecasting methodologies, is a priority for system operators and wind producers to implement low-carbon power grids. Unfortunately, though the complexity of this task implies the comparison of several alternative forecasting methodologies and dimensionality reduction techniques, a general and robust procedure of model assessment still lacks in literature. In this paper the authors propose a robust methodology, based on extensive statistical analysis and resampling routines, to supply the most effective wind power forecasting method by testing a vast ensemble of methodologies over multiple time-scales and on a real case study. Experimental results on real data collected in an Italian wind farm show the potential of ensemble approaches integrating both statistical and machine learning methods.}, note = {DOI: 10.1007/s40866-020-00090-8}, keywords = {}, pubstate = {published}, tppubtype = {article} } The massive penetration of renewable power generation in modern power grids is an effective way to reduce the impact of energy production on global warming. Unfortunately, the wind power generation may affect the regular operation of electrical systems, due to the stochastic and intermittent nature of the wind. For this reason, reducing the uncertainty about the wind evolution, e.g. by using short-term wind power forecasting methodologies, is a priority for system operators and wind producers to implement low-carbon power grids. Unfortunately, though the complexity of this task implies the comparison of several alternative forecasting methodologies and dimensionality reduction techniques, a general and robust procedure of model assessment still lacks in literature. In this paper the authors propose a robust methodology, based on extensive statistical analysis and resampling routines, to supply the most effective wind power forecasting method by testing a vast ensemble of methodologies over multiple time-scales and on a real case study. Experimental results on real data collected in an Italian wind farm show the potential of ensemble approaches integrating both statistical and machine learning methods. |
Simar, Cédric; Cebolla, Ana Maria; "e, Ga; Petieau, Mathieu; Bontempi, Gianluca; Berthoz, Alain; Cheron, Guy Hyperscanning EEG and Classification Based on Riemannian Geometry for Festive and Violent Mental State Discrimination Journal Article Frontiers in Neuroscience, 14 , 2020, (DOI: 10.3389/fnins.2020.588357). @article{info:hdl:2013/317394, title = {Hyperscanning EEG and Classification Based on Riemannian Geometry for Festive and Violent Mental State Discrimination}, author = {Cédric Simar and Ana Maria Cebolla and Ga{"e}lle Chartier and Mathieu Petieau and Gianluca Bontempi and Alain Berthoz and Guy Cheron}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/317394/3/pdf}, year = {2020}, date = {2020-01-01}, journal = {Frontiers in Neuroscience}, volume = {14}, abstract = {Interactions between two brains constitute the essence of social communication. Daily movements are commonly executed during social interactions and are determined by different mental states that may express different positive or negative behavioral intent. In this context, the effective recognition of festive or violent intent before the action execution remains crucial for survival. Here, we hypothesize that the EEG signals contain the distinctive features characterizing movement intent already expressed before movement execution and that such distinctive information can be identified by state-of-the-art classification algorithms based on Riemannian geometry. We demonstrated for the first time that a classifier based on covariance matrices and Riemannian geometry can effectively discriminate between neutral, festive, and violent mental states only on the basis of non-invasive EEG signals in both the actor and observer participants. These results pave the way for new electrophysiological discrimination of mental states based on non-invasive EEG recordings and cutting-edge machine learning techniques.}, note = {DOI: 10.3389/fnins.2020.588357}, keywords = {}, pubstate = {published}, tppubtype = {article} } Interactions between two brains constitute the essence of social communication. Daily movements are commonly executed during social interactions and are determined by different mental states that may express different positive or negative behavioral intent. In this context, the effective recognition of festive or violent intent before the action execution remains crucial for survival. Here, we hypothesize that the EEG signals contain the distinctive features characterizing movement intent already expressed before movement execution and that such distinctive information can be identified by state-of-the-art classification algorithms based on Riemannian geometry. We demonstrated for the first time that a classifier based on covariance matrices and Riemannian geometry can effectively discriminate between neutral, festive, and violent mental states only on the basis of non-invasive EEG signals in both the actor and observer participants. These results pave the way for new electrophysiological discrimination of mental states based on non-invasive EEG recordings and cutting-edge machine learning techniques. |
Colaprico, Antonio; Olsen, Catharina; Bailey, Matthew; Odom, Gabriel G J; Terkelsen, Thilde; Silva, Tiago Chedraoui; Olsen, André Vidas; Cantini, Laura; Zinovyev, Andrey; Barillot, Emmanuel; Noushmehr, Houtan; Bertoli, Gloria; Castiglioni, Isabella; Cava, Claudia; Bontempi, Gianluca; Chen, Xi Steven; Papaleo, Elena Interpreting pathways to discover cancer driver genes with Moonlight Journal Article Nature communications, 11 (1), 2020, (DOI: 10.1038/s41467-019-13803-0). @article{info:hdl:2013/301750, title = {Interpreting pathways to discover cancer driver genes with Moonlight}, author = {Antonio Colaprico and Catharina Olsen and Matthew Bailey and Gabriel G J Odom and Thilde Terkelsen and Tiago Chedraoui Silva and André Vidas Olsen and Laura Cantini and Andrey Zinovyev and Emmanuel Barillot and Houtan Noushmehr and Gloria Bertoli and Isabella Castiglioni and Claudia Cava and Gianluca Bontempi and Xi Steven Chen and Elena Papaleo}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/301750/1/doi_285394.pdf}, year = {2020}, date = {2020-01-01}, journal = {Nature communications}, volume = {11}, number = {1}, abstract = {Cancer driver gene alterations influence cancer development, occurring in oncogenes, tumor suppressors, and dual role genes. Discovering dual role cancer genes is difficult because of their elusive context-dependent behavior. We define oncogenic mediators as genes controlling biological processes. With them, we classify cancer driver genes, unveiling their roles in cancer mechanisms. To this end, we present Moonlight, a tool that incorporates multiple -omics data to identify critical cancer driver genes. With Moonlight, we analyze 8000+ tumor samples from 18 cancer types, discovering 3310 oncogenic mediators, 151 having dual roles. By incorporating additional data (amplification, mutation, DNA methylation, chromatin accessibility), we reveal 1000+ cancer driver genes, corroborating known molecular mechanisms. Additionally, we confirm critical cancer driver genes by analysing cell-line datasets. We discover inactivation of tumor suppressors in intron regions and that tissue type and subtype indicate dual role status. These findings help explain tumor heterogeneity and could guide therapeutic decisions.}, note = {DOI: 10.1038/s41467-019-13803-0}, keywords = {}, pubstate = {published}, tppubtype = {article} } Cancer driver gene alterations influence cancer development, occurring in oncogenes, tumor suppressors, and dual role genes. Discovering dual role cancer genes is difficult because of their elusive context-dependent behavior. We define oncogenic mediators as genes controlling biological processes. With them, we classify cancer driver genes, unveiling their roles in cancer mechanisms. To this end, we present Moonlight, a tool that incorporates multiple -omics data to identify critical cancer driver genes. With Moonlight, we analyze 8000+ tumor samples from 18 cancer types, discovering 3310 oncogenic mediators, 151 having dual roles. By incorporating additional data (amplification, mutation, DNA methylation, chromatin accessibility), we reveal 1000+ cancer driver genes, corroborating known molecular mechanisms. Additionally, we confirm critical cancer driver genes by analysing cell-line datasets. We discover inactivation of tumor suppressors in intron regions and that tissue type and subtype indicate dual role status. These findings help explain tumor heterogeneity and could guide therapeutic decisions. |
Duerinckx, Sarah; Jacquemin, Valérie; Drunat, Séverine; Vial, Yoann; Passemard, Sandrine; Perazzolo, Camille; Massart, Annick; Soblet, Julie; Racapé, Judith; Desmyter, Laurence; Badoer, Cindy; Papadimitriou, Sofia; "e, Yann-A; Lefort, Anne; Libert, Frédérick; Maertelaer, Viviane De; Rooman, Marianne; Costagliola, Sabine; Verloes, Alain; Lenaerts, Tom; Pirson, Isabelle; Abramowicz, Marc Digenic inheritance of human primary microcephaly delineates centrosomal and non centrosomal pathways. Journal Article Human mutation, 41 (2), pp. 512-524, 2020, (DOI: 10.1002/humu.23948). @article{info:hdl:2013/296188, title = {Digenic inheritance of human primary microcephaly delineates centrosomal and non centrosomal pathways.}, author = {Sarah Duerinckx and Valérie Jacquemin and Séverine Drunat and Yoann Vial and Sandrine Passemard and Camille Perazzolo and Annick Massart and Julie Soblet and Judith Racapé and Laurence Desmyter and Cindy Badoer and Sofia Papadimitriou and Yann-A{"e}l Le Borgne and Anne Lefort and Frédérick Libert and Viviane De Maertelaer and Marianne Rooman and Sabine Costagliola and Alain Verloes and Tom Lenaerts and Isabelle Pirson and Marc Abramowicz}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/296188/4/Supp_Mat.pdf}, year = {2020}, date = {2020-01-01}, journal = {Human mutation}, volume = {41}, number = {2}, pages = {512-524}, abstract = {Primary Microcephaly (PM) is characterized by a small head since birth and is vastly heterogeneous both genetically and phenotypically. While most cases are monogenic, genetic interactions between Aspm and Wdr62 have recently been described in a mouse model of PM. Here, we used two complementary, holistic in vivo approaches: high throughput DNA sequencing of multiple PM genes in human PM patients, and genome-edited zebrafish modeling for digenic inheritance of PM. Exomes of PM patients showed a significant burden of variants in 75 PM genes, that persisted after removing monogenic causes of PM (e.g., biallelic pathogenic variants in CEP152). This observation was replicated in an independent cohort of PM patients, where a PM gene panel showed in addition that the burden was carried by six centrosomal genes. Allelic frequencies were consistent with digenic inheritance. In zebrafish, non-centrosomal gene casc5 -/- produced a severe PM phenotype, that was not modified by centrosomal genes aspm or wdr62 invalidation. A digenic, quadriallelic PM phenotype was produced by aspm and wdr62. Our observations provide strong evidence for digenic inheritance of human PM, involving centrosomal genes. Absence of genetic interaction between casc5 and aspm or wdr62 further delineates centrosomal and non-centrosomal pathways in PM. This article is protected by copyright. All rights reserved.}, note = {DOI: 10.1002/humu.23948}, keywords = {}, pubstate = {published}, tppubtype = {article} } Primary Microcephaly (PM) is characterized by a small head since birth and is vastly heterogeneous both genetically and phenotypically. While most cases are monogenic, genetic interactions between Aspm and Wdr62 have recently been described in a mouse model of PM. Here, we used two complementary, holistic in vivo approaches: high throughput DNA sequencing of multiple PM genes in human PM patients, and genome-edited zebrafish modeling for digenic inheritance of PM. Exomes of PM patients showed a significant burden of variants in 75 PM genes, that persisted after removing monogenic causes of PM (e.g., biallelic pathogenic variants in CEP152). This observation was replicated in an independent cohort of PM patients, where a PM gene panel showed in addition that the burden was carried by six centrosomal genes. Allelic frequencies were consistent with digenic inheritance. In zebrafish, non-centrosomal gene casc5 -/- produced a severe PM phenotype, that was not modified by centrosomal genes aspm or wdr62 invalidation. A digenic, quadriallelic PM phenotype was produced by aspm and wdr62. Our observations provide strong evidence for digenic inheritance of human PM, involving centrosomal genes. Absence of genetic interaction between casc5 and aspm or wdr62 further delineates centrosomal and non-centrosomal pathways in PM. This article is protected by copyright. All rights reserved. |
Rocha, Luis Mateus; Singh, Vikramjit; Esch, Markus; Lenaerts, Tom; Liljeros, Fredrik; Thorson, Anna Dynamic contact networks of patients and MRSA spread in hospitals Journal Article Scientific reports, 10 (1), 2020, (DOI: 10.1038/s41598-020-66270-9). @article{info:hdl:2013/308993, title = {Dynamic contact networks of patients and MRSA spread in hospitals}, author = {Luis Mateus Rocha and Vikramjit Singh and Markus Esch and Tom Lenaerts and Fredrik Liljeros and Anna Thorson}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/308993/1/doi_292637.pdf}, year = {2020}, date = {2020-01-01}, journal = {Scientific reports}, volume = {10}, number = {1}, abstract = {Methicillin-resistant Staphylococcus aureus (MRSA) is a difficult-to-treat infection. Increasing efforts have been taken to mitigate the epidemics and to avoid potential outbreaks in low endemic settings. Understanding the population dynamics of MRSA is essential to identify the causal mechanisms driving the epidemics and to generalise conclusions to different contexts. Previous studies neglected the temporal structure of contacts between patients and assumed homogeneous behaviour. We developed a high-resolution data-driven contact network model of interactions between 743,182 patients in 485 hospitals during 3,059 days to reproduce the exact contact sequences of the hospital population. Our model captures the exact spatial and temporal human contact behaviour and the dynamics of referrals within and between wards and hospitals at a large scale, revealing highly heterogeneous contact and mobility patterns of individual patients. A simulation exercise of epidemic spread shows that heterogeneous contacts cause the emergence of super-spreader patients, slower than exponential polynomial growth of the prevalence, and fast epidemic spread between wards and hospitals. In our simulated scenarios, screening upon hospital admittance is potentially more effective than reducing infection probability to reduce the final outbreak size. Our findings are useful to understand not only MRSA spread but also other hospital-acquired infections.}, note = {DOI: 10.1038/s41598-020-66270-9}, keywords = {}, pubstate = {published}, tppubtype = {article} } Methicillin-resistant Staphylococcus aureus (MRSA) is a difficult-to-treat infection. Increasing efforts have been taken to mitigate the epidemics and to avoid potential outbreaks in low endemic settings. Understanding the population dynamics of MRSA is essential to identify the causal mechanisms driving the epidemics and to generalise conclusions to different contexts. Previous studies neglected the temporal structure of contacts between patients and assumed homogeneous behaviour. We developed a high-resolution data-driven contact network model of interactions between 743,182 patients in 485 hospitals during 3,059 days to reproduce the exact contact sequences of the hospital population. Our model captures the exact spatial and temporal human contact behaviour and the dynamics of referrals within and between wards and hospitals at a large scale, revealing highly heterogeneous contact and mobility patterns of individual patients. A simulation exercise of epidemic spread shows that heterogeneous contacts cause the emergence of super-spreader patients, slower than exponential polynomial growth of the prevalence, and fast epidemic spread between wards and hospitals. In our simulated scenarios, screening upon hospital admittance is potentially more effective than reducing infection probability to reduce the final outbreak size. Our findings are useful to understand not only MRSA spread but also other hospital-acquired infections. |
"e, Ma; Demetter, Pieter; Wind, Roland De; Galdon, Maria Gomez; Velde, Sylvie Vande; Lens, Gaspard; Craciun, Ligia; Deleruelle, Amélie; Larsimont, Denis; Lenaerts, Tom; Sclafani, Francesco; Deleporte, Amélie; Donckier, Vincent; Hendlisz, Alain; Vandeputte, Caroline Infiltrative tumour growth pattern correlates with poor outcome in oesophageal cancer. Journal Article BMJ open gastroenterology, 7 (1), 2020, (DOI: 10.1136/bmjgast-2020-000431). @article{info:hdl:2013/312667, title = {Infiltrative tumour growth pattern correlates with poor outcome in oesophageal cancer.}, author = {Ma{"e}lle Anciaux and Pieter Demetter and Roland De Wind and Maria Gomez Galdon and Sylvie Vande Velde and Gaspard Lens and Ligia Craciun and Amélie Deleruelle and Denis Larsimont and Tom Lenaerts and Francesco Sclafani and Amélie Deleporte and Vincent Donckier and Alain Hendlisz and Caroline Vandeputte}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/312667/1/doi_296311.pdf}, year = {2020}, date = {2020-01-01}, journal = {BMJ open gastroenterology}, volume = {7}, number = {1}, abstract = {Oesophageal cancer (OEC) is an aggressive disease with a poor survival rate. Prognostic markers are thus urgently needed. Due to the demonstrated prognostic value of histopathological growth pattern (HGP) in other cancers, we performed a retrospective assessment of HGP in patients suffering from invasive OEC.}, note = {DOI: 10.1136/bmjgast-2020-000431}, keywords = {}, pubstate = {published}, tppubtype = {article} } Oesophageal cancer (OEC) is an aggressive disease with a poor survival rate. Prognostic markers are thus urgently needed. Due to the demonstrated prognostic value of histopathological growth pattern (HGP) in other cancers, we performed a retrospective assessment of HGP in patients suffering from invasive OEC. |
'c, Jelena Gruji; Lenaerts, Tom Do people imitate when making decisions? Evidence from a spatial Prisoner's Dilemma experiment: Do people imitate when making decisions Journal Article Royal Society open science, 7 (7), 2020, (DOI: 10.1098/rsos.200618). @article{info:hdl:2013/313051, title = {Do people imitate when making decisions? Evidence from a spatial Prisoner's Dilemma experiment: Do people imitate when making decisions}, author = {Jelena Gruji{'c} and Tom Lenaerts}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/313051/3/rsos.200618.pdf}, year = {2020}, date = {2020-01-01}, journal = {Royal Society open science}, volume = {7}, number = {7}, abstract = {How do people decide which action to take? This question is best answered using Game Theory, which has proposed a series of decision-making mechanisms that people potentially use. In network simulations, wherein games are repeated and pay-off differences can be observed, those mechanisms often rely on imitation of successful behaviour. Surprisingly, little to no evidence has been provided about whether people actually imitate more successful opponents when altering their actions in that context. By comparing two experimental treatments wherein participants play the iterated Prisoner's Dilemma game in a lattice, we aim to answer whether more successful actions are imitated. While in the first treatment, participants have the possibility to use pay-off differences in making their decision, the second treatment hinders such imitation as no information about the gains is provided. If imitation of the more successful plays a role then there should be a difference in how players switch from cooperation to defection between both treatments. Although, cooperation and pay-off levels do not appear to be significantly different between both treatments, detailed analysis shows that there are behavioural differences: when confronted with a more successful co-player, the focal player will imitate her behaviour as the switching is related to the experienced pay-off inequality.}, note = {DOI: 10.1098/rsos.200618}, keywords = {}, pubstate = {published}, tppubtype = {article} } How do people decide which action to take? This question is best answered using Game Theory, which has proposed a series of decision-making mechanisms that people potentially use. In network simulations, wherein games are repeated and pay-off differences can be observed, those mechanisms often rely on imitation of successful behaviour. Surprisingly, little to no evidence has been provided about whether people actually imitate more successful opponents when altering their actions in that context. By comparing two experimental treatments wherein participants play the iterated Prisoner's Dilemma game in a lattice, we aim to answer whether more successful actions are imitated. While in the first treatment, participants have the possibility to use pay-off differences in making their decision, the second treatment hinders such imitation as no information about the gains is provided. If imitation of the more successful plays a role then there should be a difference in how players switch from cooperation to defection between both treatments. Although, cooperation and pay-off levels do not appear to be significantly different between both treatments, detailed analysis shows that there are behavioural differences: when confronted with a more successful co-player, the focal player will imitate her behaviour as the switching is related to the experienced pay-off inequality. |
Domingos, Elias Fernández; 'c, Jelena Gruji; Burguillo, Juan J C; Kirchsteiger, Georg; Santos, Francisco C; Lenaerts, Tom Timing Uncertainty in Collective Risk Dilemmas Encourages Group Reciprocation and Polarization Journal Article iScience, 23 (12), 2020, (DOI: 10.1016/j.isci.2020.101752). @article{info:hdl:2013/315240, title = {Timing Uncertainty in Collective Risk Dilemmas Encourages Group Reciprocation and Polarization}, author = {Elias Fernández Domingos and Jelena Gruji{'c} and Juan J C Burguillo and Georg Kirchsteiger and Francisco C Santos and Tom Lenaerts}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/315240/1/doi_298884.pdf}, year = {2020}, date = {2020-01-01}, journal = {iScience}, volume = {23}, number = {12}, abstract = {Social dilemmas are often shaped by actions involving uncertain returns only achievable in the future, such as climate action or voluntary vaccination. In this context, uncertainty may produce non-trivial effects. Here, we assess experimentally --- through a collective risk dilemma --- the effect of timing uncertainty, i.e. how uncertainty about when a target needs to be reached affects the participants' behaviors. We show that timing uncertainty prompts not only early generosity but also polarized outcomes, where participants' total contributions are distributed unevenly. Furthermore, analyzing participants' behavior under timing uncertainty reveals an increase in reciprocal strategies. A data-driven game-theoretical model captures the self-organizing dynamics underpinning these behavioral patterns. Timing uncertainty thus casts a shadow on the future that leads participants to respond early, whereas reciprocal strategies appear to be important for group success. Yet, the same uncertainty also leads to inequity and polarization, requiring the inclusion of new incentives handling these societal issues.}, note = {DOI: 10.1016/j.isci.2020.101752}, keywords = {}, pubstate = {published}, tppubtype = {article} } Social dilemmas are often shaped by actions involving uncertain returns only achievable in the future, such as climate action or voluntary vaccination. In this context, uncertainty may produce non-trivial effects. Here, we assess experimentally --- through a collective risk dilemma --- the effect of timing uncertainty, i.e. how uncertainty about when a target needs to be reached affects the participants' behaviors. We show that timing uncertainty prompts not only early generosity but also polarized outcomes, where participants' total contributions are distributed unevenly. Furthermore, analyzing participants' behavior under timing uncertainty reveals an increase in reciprocal strategies. A data-driven game-theoretical model captures the self-organizing dynamics underpinning these behavioral patterns. Timing uncertainty thus casts a shadow on the future that leads participants to respond early, whereas reciprocal strategies appear to be important for group success. Yet, the same uncertainty also leads to inequity and polarization, requiring the inclusion of new incentives handling these societal issues. |
Abels, Axel; Lenaerts, Tom; Trianni, Vito; Nowe, Ann How Expert Confidence Can Improve Collective Decision-Making in Contextual Multi-Armed Bandit Problems Journal Article Lecture notes in computer science, 12496 LNAI , pp. 125-138, 2020, (DOI: 10.1007/978-3-030-63007-2_10). @article{info:hdl:2013/316598, title = {How Expert Confidence Can Improve Collective Decision-Making in Contextual Multi-Armed Bandit Problems}, author = {Axel Abels and Tom Lenaerts and Vito Trianni and Ann Nowe}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/316598/4/howexpertconfidence.pdf}, year = {2020}, date = {2020-01-01}, journal = {Lecture notes in computer science}, volume = {12496 LNAI}, pages = {125-138}, abstract = {In collective decision-making (CDM) a group of experts with a shared set of values and a common goal must combine their knowledge to make a collectively optimal decision. Whereas existing research on CDM primarily focuses on making binary decisions, we focus here on CDM applied to solving contextual multi-armed bandit (CMAB) problems, where the goal is to exploit contextual information to select the best arm among a set. To address the limiting assumptions of prior work, we introduce confidence estimates and propose a novel approach to deciding with expert advice which can take advantage of these estimates. We further show that, when confidence estimates are imperfect, the proposed approach is more robust than the classical confidence-weighted majority vote.}, note = {DOI: 10.1007/978-3-030-63007-2_10}, keywords = {}, pubstate = {published}, tppubtype = {article} } In collective decision-making (CDM) a group of experts with a shared set of values and a common goal must combine their knowledge to make a collectively optimal decision. Whereas existing research on CDM primarily focuses on making binary decisions, we focus here on CDM applied to solving contextual multi-armed bandit (CMAB) problems, where the goal is to exploit contextual information to select the best arm among a set. To address the limiting assumptions of prior work, we introduce confidence estimates and propose a novel approach to deciding with expert advice which can take advantage of these estimates. We further show that, when confidence estimates are imperfect, the proposed approach is more robust than the classical confidence-weighted majority vote. |
Abels, Axel; Lenaerts, Tom; Trianni, Vito; Nowe, Ann Collective Decision-Making as a Contextual Multi-armed Bandit Problem Journal Article Lecture notes in computer science, 12496 LNAI , pp. 113-124, 2020, (DOI: 10.1007/978-3-030-63007-2_9). @article{info:hdl:2013/316605, title = {Collective Decision-Making as a Contextual Multi-armed Bandit Problem}, author = {Axel Abels and Tom Lenaerts and Vito Trianni and Ann Nowe}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/316605/5/colascdm.pdf}, year = {2020}, date = {2020-01-01}, journal = {Lecture notes in computer science}, volume = {12496 LNAI}, pages = {113-124}, abstract = {Collective decision-making (CDM) processes -- wherein the knowledge of a group of individuals with a common goal must be combined to make optimal decisions -- can be formalized within the framework of the deciding with expert advice setting. Traditional approaches to tackle this problem focus on finding appropriate weights for the individuals in the group. In contrast, we propose here meta-CMAB, a meta approach that learns a mapping from expert advice to expected outcomes. In summary, our work reveals that, when trying to make the best choice in a problem with multiple alternatives, meta-CMAB assures that the collective knowledge of experts leads to the best outcome without the need for accurate confidence estimates.}, note = {DOI: 10.1007/978-3-030-63007-2_9}, keywords = {}, pubstate = {published}, tppubtype = {article} } Collective decision-making (CDM) processes -- wherein the knowledge of a group of individuals with a common goal must be combined to make optimal decisions -- can be formalized within the framework of the deciding with expert advice setting. Traditional approaches to tackle this problem focus on finding appropriate weights for the individuals in the group. In contrast, we propose here meta-CMAB, a meta approach that learns a mapping from expert advice to expected outcomes. In summary, our work reveals that, when trying to make the best choice in a problem with multiple alternatives, meta-CMAB assures that the collective knowledge of experts leads to the best outcome without the need for accurate confidence estimates. |
Han, The Anh T A H; 'i, Lu; Santos, Francisco C; Lenaerts, Tom To regulate or not: A social dynamics analysis of an idealised ai race Journal Article The journal of artificial intelligence research, 69 , pp. 881-921, 2020, (DOI: 10.1613/JAIR.1.12225). @article{info:hdl:2013/319033, title = {To regulate or not: A social dynamics analysis of an idealised ai race}, author = {The Anh T A H Han and Lu{'i}s Marcelo Pereira and Francisco C Santos and Tom Lenaerts}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/319033/17/1907-12393.pdf}, year = {2020}, date = {2020-01-01}, journal = {The journal of artificial intelligence research}, volume = {69}, pages = {881-921}, abstract = {Rapid technological advancements in Artificial Intelligence (AI), as well as the growing deployment of intelligent technologies in new application domains, have generated serious anxiety and a fear of missing out among different stake-holders, fostering a racing narrative. Whether real or not, the belief in such a race for domain supremacy through AI, can make it real simply from its consequences, as put forward by the Thomas theorem. These consequences may be negative, as racing for technological supremacy creates a complex ecology of choices that could push stake-holders to underestimate or even ignore ethical and safety procedures. As a consequence, different actors are urging to consider both the normative and social impact of these technological advancements, contemplating the use of the precautionary principle in AI innovation and research. Yet, given the breadth and depth of AI and its advances, it is difficult to assess which technology needs regulation and when. As there is no easy access to data describing this alleged AI race, theoretical models are necessary to understand its potential dynamics, allowing for the identification of when procedures need to be put in place to favour outcomes beneficial for all. We show that, next to the risks of setbacks and being reprimanded for unsafe behaviour, the time-scale in which domain supremacy can be achieved plays a crucial role. When this can be achieved in a short term, those who completely ignore the safety precautions are bound to win the race but at a cost to society, apparently requiring regulatory actions. Our analysis reveals that imposing regulations for all risk and timing conditions may not have the anticipated effect as only for specific conditions a dilemma arises between what is individually preferred and globally beneficial. Similar observations can be made for the long-term development case. Yet different from the short-term situation, conditions can be identified that require the promotion of risk-taking as opposed to compliance with safety regulations in order to improve social welfare. These results remain robust both when two or several actors are involved in the race and when collective rather than individual setbacks are produced by risk-taking behaviour. When defining codes of conduct and regulatory policies for applications of AI, a clear understanding of the time-scale of the race is thus required, as this may induce important non-trivial effects.}, note = {DOI: 10.1613/JAIR.1.12225}, keywords = {}, pubstate = {published}, tppubtype = {article} } Rapid technological advancements in Artificial Intelligence (AI), as well as the growing deployment of intelligent technologies in new application domains, have generated serious anxiety and a fear of missing out among different stake-holders, fostering a racing narrative. Whether real or not, the belief in such a race for domain supremacy through AI, can make it real simply from its consequences, as put forward by the Thomas theorem. These consequences may be negative, as racing for technological supremacy creates a complex ecology of choices that could push stake-holders to underestimate or even ignore ethical and safety procedures. As a consequence, different actors are urging to consider both the normative and social impact of these technological advancements, contemplating the use of the precautionary principle in AI innovation and research. Yet, given the breadth and depth of AI and its advances, it is difficult to assess which technology needs regulation and when. As there is no easy access to data describing this alleged AI race, theoretical models are necessary to understand its potential dynamics, allowing for the identification of when procedures need to be put in place to favour outcomes beneficial for all. We show that, next to the risks of setbacks and being reprimanded for unsafe behaviour, the time-scale in which domain supremacy can be achieved plays a crucial role. When this can be achieved in a short term, those who completely ignore the safety precautions are bound to win the race but at a cost to society, apparently requiring regulatory actions. Our analysis reveals that imposing regulations for all risk and timing conditions may not have the anticipated effect as only for specific conditions a dilemma arises between what is individually preferred and globally beneficial. Similar observations can be made for the long-term development case. Yet different from the short-term situation, conditions can be identified that require the promotion of risk-taking as opposed to compliance with safety regulations in order to improve social welfare. These results remain robust both when two or several actors are involved in the race and when collective rather than individual setbacks are produced by risk-taking behaviour. When defining codes of conduct and regulatory policies for applications of AI, a clear understanding of the time-scale of the race is thus required, as this may induce important non-trivial effects. |
Lorenzo, Ramiro; Onizuka, Michiho; Defrance, Matthieu; Laurent, Patrick Combining single-cell RNA-sequencing with a molecular atlas unveils new markers for Caenorhabditis elegans neuron classes Journal Article Nucleic acids research, 2020, (DOI: 10.1093/nar/gkaa486). @article{info:hdl:2013/316017, title = {Combining single-cell RNA-sequencing with a molecular atlas unveils new markers for Caenorhabditis elegans neuron classes}, author = {Ramiro Lorenzo and Michiho Onizuka and Matthieu Defrance and Patrick Laurent}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/316017/1/doi_299661.pdf}, year = {2020}, date = {2020-01-01}, journal = {Nucleic acids research}, abstract = {Abstract Single-cell RNA-sequencing (scRNA-seq) of the Caenorhabditis elegans nervous system offers the unique opportunity to obtain a partial expression profile for each neuron within a known connectome. Building on recent scRNA-seq data and on a molecular atlas describing the expression pattern of ∼800 genes at the single cell resolution, we designed an iterative clustering analysis aiming to match each cell-cluster to the ∼100 anatomically defined neuron classes of C. elegans. This heuristic approach successfully assigned 97 of the 118 neuron classes to a cluster. Sixty two clusters were assigned to a single neuron class and 15 clusters grouped neuron classes sharing close molecular signatures. Pseudotime analysis revealed a maturation process occurring in some neurons (e.g. PDA) during the L2 stage. Based on the molecular profiles of all identified neurons, we predicted cell fate regulators and experimentally validated unc-86 for the normal differentiation of RMG neurons. Furthermore, we observed that different classes of genes functionally diversify sensory neurons, interneurons and motorneurons. Finally, we designed 15 new neuron class-specific promoters validated in vivo. Amongst them, 10 represent the only specific promoter reported to this day, expanding the list of neurons amenable to genetic manipulations.}, note = {DOI: 10.1093/nar/gkaa486}, keywords = {}, pubstate = {published}, tppubtype = {article} } Abstract Single-cell RNA-sequencing (scRNA-seq) of the Caenorhabditis elegans nervous system offers the unique opportunity to obtain a partial expression profile for each neuron within a known connectome. Building on recent scRNA-seq data and on a molecular atlas describing the expression pattern of ∼800 genes at the single cell resolution, we designed an iterative clustering analysis aiming to match each cell-cluster to the ∼100 anatomically defined neuron classes of C. elegans. This heuristic approach successfully assigned 97 of the 118 neuron classes to a cluster. Sixty two clusters were assigned to a single neuron class and 15 clusters grouped neuron classes sharing close molecular signatures. Pseudotime analysis revealed a maturation process occurring in some neurons (e.g. PDA) during the L2 stage. Based on the molecular profiles of all identified neurons, we predicted cell fate regulators and experimentally validated unc-86 for the normal differentiation of RMG neurons. Furthermore, we observed that different classes of genes functionally diversify sensory neurons, interneurons and motorneurons. Finally, we designed 15 new neuron class-specific promoters validated in vivo. Amongst them, 10 represent the only specific promoter reported to this day, expanding the list of neurons amenable to genetic manipulations. |
Lipski, D; Foucart, Vincent; Dewispelaere, Remi; Caspers, Laure; Defrance, Matthieu; Bruyns, Catherine; Willermain, Francois Retinal endothelial cell phenotypic modifications during experimental autoimmune uveitis: A transcriptomic approach Journal Article BMC ophthalmology, 20 (1), 2020, (DOI: 10.1186/s12886-020-1333-5). @article{info:hdl:2013/305060, title = {Retinal endothelial cell phenotypic modifications during experimental autoimmune uveitis: A transcriptomic approach}, author = {D Lipski and Vincent Foucart and Remi Dewispelaere and Laure Caspers and Matthieu Defrance and Catherine Bruyns and Francois Willermain}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/305060/1/doi_288704.pdf}, year = {2020}, date = {2020-01-01}, journal = {BMC ophthalmology}, volume = {20}, number = {1}, abstract = {Background: Blood-retinal barrier cells are known to exhibit a massive phenotypic change during experimental autoimmune uveitis (EAU) development. In an attempt to investigate the mechanisms of blood-retinal barrier (BRB) breakdown at a global level, we studied the gene regulation of total retinal cells and retinal endothelial cells during non-infectious uveitis. Methods: Retinal endothelial cells were isolated by flow cytometry either in Tie2-GFP mice (CD31+ CD45- GFP+ cells), or in wild type C57BL/6 mice (CD31+ CD45- endoglin+ cells). EAU was induced in C57BL/6 mice by adoptive transfer of IRBP1-20-specific T cells. Total retinal cells and retinal endothelial cells from na"ive and EAU mice were sorted and their gene expression compared by RNA-Seq. Protein expression of selected genes was validated by immunofluorescence on retinal wholemounts and cryosections and by flow cytometry. Results: Retinal endothelial cell sorting in wild type C57BL/6 mice was validated by comparative transcriptome analysis with retinal endothelial cells sorted from Tie2-GFP mice, which express GFP under the control of the endothelial-specific receptor tyrosine kinase promoter Tie2. RNA-Seq analysis of total retinal cells mainly brought to light upregulation of genes involved in antigen presentation and T cell activation during EAU. Specific transcriptome analysis of retinal endothelial cells allowed us to identify 82 genes modulated in retinal endothelial cells during EAU development. Protein expression of 5 of those genes (serpina3n, lcn2, ackr1, lrg1 and lamc3) was validated at the level of inner BRB cells. Conclusion: Those data not only confirm the involvement of known pathogenic molecules but further provide a list of new candidate genes and pathways possibly implicated in inner BRB breakdown during non-infectious posterior uveitis.}, note = {DOI: 10.1186/s12886-020-1333-5}, keywords = {}, pubstate = {published}, tppubtype = {article} } Background: Blood-retinal barrier cells are known to exhibit a massive phenotypic change during experimental autoimmune uveitis (EAU) development. In an attempt to investigate the mechanisms of blood-retinal barrier (BRB) breakdown at a global level, we studied the gene regulation of total retinal cells and retinal endothelial cells during non-infectious uveitis. Methods: Retinal endothelial cells were isolated by flow cytometry either in Tie2-GFP mice (CD31+ CD45- GFP+ cells), or in wild type C57BL/6 mice (CD31+ CD45- endoglin+ cells). EAU was induced in C57BL/6 mice by adoptive transfer of IRBP1-20-specific T cells. Total retinal cells and retinal endothelial cells from na"ive and EAU mice were sorted and their gene expression compared by RNA-Seq. Protein expression of selected genes was validated by immunofluorescence on retinal wholemounts and cryosections and by flow cytometry. Results: Retinal endothelial cell sorting in wild type C57BL/6 mice was validated by comparative transcriptome analysis with retinal endothelial cells sorted from Tie2-GFP mice, which express GFP under the control of the endothelial-specific receptor tyrosine kinase promoter Tie2. RNA-Seq analysis of total retinal cells mainly brought to light upregulation of genes involved in antigen presentation and T cell activation during EAU. Specific transcriptome analysis of retinal endothelial cells allowed us to identify 82 genes modulated in retinal endothelial cells during EAU development. Protein expression of 5 of those genes (serpina3n, lcn2, ackr1, lrg1 and lamc3) was validated at the level of inner BRB cells. Conclusion: Those data not only confirm the involvement of known pathogenic molecules but further provide a list of new candidate genes and pathways possibly implicated in inner BRB breakdown during non-infectious posterior uveitis. |
2019 |
Stefani, Jacopo De; Bontempi, Gianluca; Caelen, Olivier; Hattab, Dalila SYSTEM AND METHOD FOR MANAGING RISKS IN A PROCESS Miscellaneous 2019, (Language of publication: fr). @misc{info:hdl:2013/283233, title = {SYSTEM AND METHOD FOR MANAGING RISKS IN A PROCESS}, author = {Jacopo De Stefani and Gianluca Bontempi and Olivier Caelen and Dalila Hattab}, url = {http://hdl.handle.net/2013/ULB-DIPOT:oai:dipot.ulb.ac.be:2013/283233}, year = {2019}, date = {2019-01-01}, abstract = {The invention relates to a system and a method for managing risk in a process of an industrial production chain, said system comprising at least one database containing at least one set of time series of raw data relating to said process, a set of devices storing at least one module or a set of modules, the execution of which makes it possible to implement the risk management method comprising at least the preprocessing of a set of raw data so as to produce a set of "proxies" of volatility, the analysis of correlations between the "proxies", the identification of at least one model and of optimal parameters depending at least on said "proxies" and the calculation of the predictions of the set of the "proxies" comprising the combination of at least one of said "proxies" with at least one exogenous parameter included in said set of "proxies".}, note = {Language of publication: fr}, keywords = {}, pubstate = {published}, tppubtype = {misc} } The invention relates to a system and a method for managing risk in a process of an industrial production chain, said system comprising at least one database containing at least one set of time series of raw data relating to said process, a set of devices storing at least one module or a set of modules, the execution of which makes it possible to implement the risk management method comprising at least the preprocessing of a set of raw data so as to produce a set of "proxies" of volatility, the analysis of correlations between the "proxies", the identification of at least one model and of optimal parameters depending at least on said "proxies" and the calculation of the predictions of the set of the "proxies" comprising the combination of at least one of said "proxies" with at least one exogenous parameter included in said set of "proxies". |
Stefani, Jacopo De; Caelen, Olivier; Hattab, Dalila; "e, Yann-A; Bontempi, Gianluca A Multivariate and Multi-step Ahead Machine Learning Approach to Traditional and Cryptocurrencies Volatility Forecasting Inproceedings ECML PKDD 2018 Workshops, Springer, 2019, (Conference: ECML-PKDD 2018(Dublin)). @inproceedings{info:hdl:2013/284007, title = {A Multivariate and Multi-step Ahead Machine Learning Approach to Traditional and Cryptocurrencies Volatility Forecasting}, author = {Jacopo De Stefani and Olivier Caelen and Dalila Hattab and Yann-A{"e}l Le Borgne and Gianluca Bontempi}, url = {http://hdl.handle.net/2013/ULB-DIPOT:oai:dipot.ulb.ac.be:2013/284007}, year = {2019}, date = {2019-01-01}, booktitle = {ECML PKDD 2018 Workshops}, publisher = {Springer}, series = {Lecture Notes in Computer Science, 11054}, abstract = {Multivariate time series forecasting involves the learning of historical multivariate information in order to predict the future values of several quantities of interests, accounting for interdependencies among them. In finance, several of this quantities of interests (stock valuations, return, volatility) have been shown to be mutually influencing each other, making the prediction of such quantities a difficult task, especially while dealing with an high number of variables and multiple horizons in the future. Here we propose a machine learning based framework, the DFML, based on the Dynamic Factor Model, to first perform a dimensionality reduction and then perform a multiple step ahead forecasting of a reduced number of components. Finally, the components are transformed again into an high dimensional space, providing the desired forecast. Our results, comparing the DFML with several state of the art techniques from different domanins (PLS, RNN, LSTM, DFM), on both traditional stock markets and cryptocurrencies market and for different families of volatility proxies show that the DFML outperforms the concurrent methods, especially for longer horizons. We conclude by explaining how we wish to further improve the performances of the framework, both in terms of accuracy and computational efficiency.}, note = {Conference: ECML-PKDD 2018(Dublin)}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } Multivariate time series forecasting involves the learning of historical multivariate information in order to predict the future values of several quantities of interests, accounting for interdependencies among them. In finance, several of this quantities of interests (stock valuations, return, volatility) have been shown to be mutually influencing each other, making the prediction of such quantities a difficult task, especially while dealing with an high number of variables and multiple horizons in the future. Here we propose a machine learning based framework, the DFML, based on the Dynamic Factor Model, to first perform a dimensionality reduction and then perform a multiple step ahead forecasting of a reduced number of components. Finally, the components are transformed again into an high dimensional space, providing the desired forecast. Our results, comparing the DFML with several state of the art techniques from different domanins (PLS, RNN, LSTM, DFM), on both traditional stock markets and cryptocurrencies market and for different families of volatility proxies show that the DFML outperforms the concurrent methods, especially for longer horizons. We conclude by explaining how we wish to further improve the performances of the framework, both in terms of accuracy and computational efficiency. |
Verhelst, Theo; Caelen, Olivier; Dewitte, Jean Christophe; Lebichot, Bertrand; Bontempi, Gianluca Understanding telecom customer churn with machine learning: From prediction to causal inference Journal Article CEUR Workshop Proceedings, 2491 , 2019, (Language of publication: en). @article{info:hdl:2013/296511, title = {Understanding telecom customer churn with machine learning: From prediction to causal inference}, author = {Theo Verhelst and Olivier Caelen and Jean Christophe Dewitte and Bertrand Lebichot and Gianluca Bontempi}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/296511/3/abstract28.pdf}, year = {2019}, date = {2019-01-01}, journal = {CEUR Workshop Proceedings}, volume = {2491}, note = {Language of publication: en}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
Mounir, Mohamed; Lucchetta, Marta; Silva, Tiago Henrique Da T C; Olsen, Catharina; Bontempi, Gianluca; Chen, Xi; Noushmehr, Houtan; Colaprico, Antonio; Papaleo, Elena New functionalities in the TCGAbiolinks package for the study and integration of cancer data from GDC and GTEx Journal Article PLoS computational biology, 15 (3), pp. e1006701, 2019, (DOI: 10.1371/journal.pcbi.1006701). @article{info:hdl:2013/286948, title = {New functionalities in the TCGAbiolinks package for the study and integration of cancer data from GDC and GTEx}, author = {Mohamed Mounir and Marta Lucchetta and Tiago Henrique Da T C Silva and Catharina Olsen and Gianluca Bontempi and Xi Chen and Houtan Noushmehr and Antonio Colaprico and Elena Papaleo}, url = {http://hdl.handle.net/2013/ULB-DIPOT:oai:dipot.ulb.ac.be:2013/286948}, year = {2019}, date = {2019-01-01}, journal = {PLoS computational biology}, volume = {15}, number = {3}, pages = {e1006701}, abstract = {The advent of Next-Generation Sequencing (NGS) technologies has opened new perspectives in deciphering the genetic mechanisms underlying complex diseases. Nowadays, the amount of genomic data is massive and substantial efforts and new tools are required to unveil the information hidden in the data. The Genomic Data Commons (GDC) Data Portal is a platform that contains different genomic studies including the ones from The Cancer Genome Atlas (TCGA) and the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) initiatives, accounting for more than 40 tumor types originating from nearly 30000 patients. Such platforms, although very attractive, must make sure the stored data are easily accessible and adequately harmonized. Moreover, they have the primary focus on the data storage in a unique place, and they do not provide a comprehensive toolkit for analyses and interpretation of the data. To fulfill this urgent need, comprehensive but easily accessible computational methods for integrative analyses of genomic data that do not renounce a robust statistical and theoretical framework are required. In this context, the R/Bioconductor package TCGAbiolinks was developed, offering a variety of bioinformatics functionalities. Here we introduce new features and enhancements of TCGAbiolinks in terms of i) more accurate and flexible pipelines for differential expression analyses, ii) different methods for tumor purity estimation and filtering, iii) integration of normal samples from other platforms iv) support for other genomics datasets, exemplified here by the TARGET data. Evidence has shown that accounting for tumor purity is essential in the study of tumorigenesis, as these factors promote confounding behavior regarding differential expression analysis. With this in mind, we implemented these filtering procedures in TCGAbiolinks. Moreover, a limitation of some of the TCGA datasets is the unavailability or paucity of corresponding normal samples. We thus integrated into TCGAbiolinks the possibility to use normal samples from the Genotype-Tissue Expression (GTEx) project, which is another large-scale repository cataloging gene expression from healthy individuals. The new functionalities are available in the TCGAbiolinks version 2.8 and higher released in Bioconductor version 3.7.}, note = {DOI: 10.1371/journal.pcbi.1006701}, keywords = {}, pubstate = {published}, tppubtype = {article} } The advent of Next-Generation Sequencing (NGS) technologies has opened new perspectives in deciphering the genetic mechanisms underlying complex diseases. Nowadays, the amount of genomic data is massive and substantial efforts and new tools are required to unveil the information hidden in the data. The Genomic Data Commons (GDC) Data Portal is a platform that contains different genomic studies including the ones from The Cancer Genome Atlas (TCGA) and the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) initiatives, accounting for more than 40 tumor types originating from nearly 30000 patients. Such platforms, although very attractive, must make sure the stored data are easily accessible and adequately harmonized. Moreover, they have the primary focus on the data storage in a unique place, and they do not provide a comprehensive toolkit for analyses and interpretation of the data. To fulfill this urgent need, comprehensive but easily accessible computational methods for integrative analyses of genomic data that do not renounce a robust statistical and theoretical framework are required. In this context, the R/Bioconductor package TCGAbiolinks was developed, offering a variety of bioinformatics functionalities. Here we introduce new features and enhancements of TCGAbiolinks in terms of i) more accurate and flexible pipelines for differential expression analyses, ii) different methods for tumor purity estimation and filtering, iii) integration of normal samples from other platforms iv) support for other genomics datasets, exemplified here by the TARGET data. Evidence has shown that accounting for tumor purity is essential in the study of tumorigenesis, as these factors promote confounding behavior regarding differential expression analysis. With this in mind, we implemented these filtering procedures in TCGAbiolinks. Moreover, a limitation of some of the TCGA datasets is the unavailability or paucity of corresponding normal samples. We thus integrated into TCGAbiolinks the possibility to use normal samples from the Genotype-Tissue Expression (GTEx) project, which is another large-scale repository cataloging gene expression from healthy individuals. The new functionalities are available in the TCGAbiolinks version 2.8 and higher released in Bioconductor version 3.7. |
Bontempi, Gianluca Comments on M4 competition Journal Article International journal of forecasting, 2019, (DOI: 10.1016/j.ijforecast.2019.03.028). @article{info:hdl:2013/292419, title = {Comments on M4 competition}, author = {Gianluca Bontempi}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/292419/1/Elsevier_276046.pdf}, year = {2019}, date = {2019-01-01}, journal = {International journal of forecasting}, note = {DOI: 10.1016/j.ijforecast.2019.03.028}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
Carcillo, Fabrizio; "e, Yann-A; Caelen, Olivier; Kessaci, Yacine; Oblé, Frédéric; Bontempi, Gianluca Combining unsupervised and supervised learning in credit card fraud detection Journal Article Information sciences, 2019, (DOI: 10.1016/j.ins.2019.05.042). @article{info:hdl:2013/289125, title = {Combining unsupervised and supervised learning in credit card fraud detection}, author = {Fabrizio Carcillo and Yann-A{"e}l Le Borgne and Olivier Caelen and Yacine Kessaci and Frédéric Oblé and Gianluca Bontempi}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/289125/1/Elsevier_272752.pdf}, year = {2019}, date = {2019-01-01}, journal = {Information sciences}, abstract = {Supervised learning techniques are widely employed in credit card fraud detection, as they make use of the assumption that fraudulent patterns can be learned from an analysis of past transactions. The task becomes challenging, however, when it has to take account of changes in customer behavior and fraudsters' ability to invent novel fraud patterns. In this context, unsupervised learning techniques can help the fraud detection systems to find anomalies. In this paper we present a hybrid technique that combines supervised and unsupervised techniques to improve the fraud detection accuracy. Unsupervised outlier scores, computed at different levels of granularity, are compared and tested on a real, annotated, credit card fraud detection dataset. Experimental results show that the combination is efficient and does indeed improve the accuracy of the detection.}, note = {DOI: 10.1016/j.ins.2019.05.042}, keywords = {}, pubstate = {published}, tppubtype = {article} } Supervised learning techniques are widely employed in credit card fraud detection, as they make use of the assumption that fraudulent patterns can be learned from an analysis of past transactions. The task becomes challenging, however, when it has to take account of changes in customer behavior and fraudsters' ability to invent novel fraud patterns. In this context, unsupervised learning techniques can help the fraud detection systems to find anomalies. In this paper we present a hybrid technique that combines supervised and unsupervised techniques to improve the fraud detection accuracy. Unsupervised outlier scores, computed at different levels of granularity, are compared and tested on a real, annotated, credit card fraud detection dataset. Experimental results show that the combination is efficient and does indeed improve the accuracy of the detection. |
Bontempi, Gianluca The Induction Problem: A Machine Learning Vindication Argument Journal Article Lecture notes in computer science, 11943 LNCS , pp. 232-243, 2019, (DOI: 10.1007/978-3-030-37599-7_20). @article{info:hdl:2013/303320, title = {The Induction Problem: A Machine Learning Vindication Argument}, author = {Gianluca Bontempi}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/303320/3/lod2.pdf}, year = {2019}, date = {2019-01-01}, journal = {Lecture notes in computer science}, volume = {11943 LNCS}, pages = {232-243}, abstract = {The problem of induction is a central problem in philosophy of science and concerns whether it is sound or not to extract laws from observational data. Nowadays, this issue is more relevant than ever given the pervasive and growing role of the data discovery process in all sciences. If on one hand induction is routinely employed by automatic machine learning techniques, on the other most of the philosophical work criticises induction as if an alternative could exist. But is there indeed a reliable alternative to induction? Is it possible to discover or predict something in a non inductive manner? This paper formalises the question on the basis of statistical notions (bias, variance, mean squared error) borrowed from estimation theory and statistical machine learning. The result is a justification of induction as rational behaviour. In a decision-making process a behaviour is rational if it is based on making choices that result in the most optimal level of benefit or utility. If we measure utility in a prediction context in terms of expected accuracy, it follows that induction is the rational way of conduct.}, note = {DOI: 10.1007/978-3-030-37599-7_20}, keywords = {}, pubstate = {published}, tppubtype = {article} } The problem of induction is a central problem in philosophy of science and concerns whether it is sound or not to extract laws from observational data. Nowadays, this issue is more relevant than ever given the pervasive and growing role of the data discovery process in all sciences. If on one hand induction is routinely employed by automatic machine learning techniques, on the other most of the philosophical work criticises induction as if an alternative could exist. But is there indeed a reliable alternative to induction? Is it possible to discover or predict something in a non inductive manner? This paper formalises the question on the basis of statistical notions (bias, variance, mean squared error) borrowed from estimation theory and statistical machine learning. The result is a justification of induction as rational behaviour. In a decision-making process a behaviour is rational if it is based on making choices that result in the most optimal level of benefit or utility. If we measure utility in a prediction context in terms of expected accuracy, it follows that induction is the rational way of conduct. |
Abels, Axel; Roijers, Diederik D M; Lenaerts, Tom; Nowe, Ann; Steckelmacher, Denis Dynamic Weights in Multi-Objective Deep Reinforcement Learning Inproceedings Proceedings of the 36th International Conference on Machine Learning, pp. 11-20, PMLR, 2019, (Language of publication: en). @inproceedings{info:hdl:2013/291979, title = {Dynamic Weights in Multi-Objective Deep Reinforcement Learning}, author = {Axel Abels and Diederik D M Roijers and Tom Lenaerts and Ann Nowe and Denis Steckelmacher}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/291979/3/abels19a.pdf}, year = {2019}, date = {2019-01-01}, booktitle = {Proceedings of the 36th International Conference on Machine Learning}, pages = {11-20}, publisher = {PMLR}, series = {Proceedings of Machine Learning Research}, note = {Language of publication: en}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
Coppens, Youri; Efthymiadis, Kyriakos; Lenaerts, Tom; Nowé, Ann Distilling Deep Reinforcement Learning Policies in Soft Decision Trees Inproceedings Miller, Tim; Weber, Rosina; Magazzeni, Daniele (Ed.): Proceedings of the IJCAI 2019 Workshop on Explainable Artificial Intelligence, 2019, (Conference: (Macau, China)). @inproceedings{info:hdl:2013/302065, title = {Distilling Deep Reinforcement Learning Policies in Soft Decision Trees}, author = {Youri Coppens and Kyriakos Efthymiadis and Tom Lenaerts and Ann Nowé}, editor = {Tim Miller and Rosina Weber and Daniele Magazzeni}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/302065/3/xrl.pdf}, year = {2019}, date = {2019-01-01}, booktitle = {Proceedings of the IJCAI 2019 Workshop on Explainable Artificial Intelligence}, abstract = {An important step in Reinforcement Learning (RL) research is to create mechanisms that give higher level insights into the black-box policy models used nowadays and provide explanations for these learned behaviors or motivate the choices behind certain decision steps. In this paper, we illustrate how Soft Decision Tree (SDT) distillation can be used to make policies that are learned through RL more interpretable. Soft Decision Trees create binary trees of predetermined depth, where each branching node represents a hierarchical filter that influences the classification of input data. We distill SDTs from a deep neural network RL policy for the Mario AI benchmark and inspect the learned hierarchy of filters, showing which input features lead to specific action distributions in the episode. We realize preliminary steps towards interpreting the learned behavior of the policy and discuss future improvements.}, note = {Conference: (Macau, China)}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } An important step in Reinforcement Learning (RL) research is to create mechanisms that give higher level insights into the black-box policy models used nowadays and provide explanations for these learned behaviors or motivate the choices behind certain decision steps. In this paper, we illustrate how Soft Decision Tree (SDT) distillation can be used to make policies that are learned through RL more interpretable. Soft Decision Trees create binary trees of predetermined depth, where each branching node represents a hierarchical filter that influences the classification of input data. We distill SDTs from a deep neural network RL policy for the Mario AI benchmark and inspect the learned hierarchy of filters, showing which input features lead to specific action distributions in the episode. We realize preliminary steps towards interpreting the learned behavior of the policy and discuss future improvements. |
Han, The Anh T A H; 'i, Lu; Lenaerts, Tom Modelling and influencing the AI bidding War: A research agenda Inproceedings Proceedings of the 2019 AAAI/ACM Conference on AI, Ethics, and Society, Association for Computing Machinery, 2019, (Conference: AAAI/ACM Conference on AI, Ethics, and Society(Honolulu, HI, USA)). @inproceedings{info:hdl:2013/314481, title = {Modelling and influencing the AI bidding War: A research agenda}, author = {The Anh T A H Han and Lu{'i}s Moniz Pereira and Tom Lenaerts}, url = {http://hdl.handle.net/2013/ULB-DIPOT:oai:dipot.ulb.ac.be:2013/314481}, year = {2019}, date = {2019-01-01}, booktitle = {Proceedings of the 2019 AAAI/ACM Conference on AI, Ethics, and Society}, publisher = {Association for Computing Machinery}, series = {AIES'19 series}, abstract = {A race for technological supremacy in AI could lead to serious negative consequences, especially whenever ethical and safety procedures are underestimated or even ignored, leading potentially to the rejection of AI in general. For all to enjoy the benefits provided by safe, ethical and trustworthy AI systems, it is crucial to incentivise participants with appropriate strategies that ensure mutually beneficial normative behaviour and safety-compliance from all parties involved. Little attention has been given to understanding the dynamics and emergent behaviours arising from this AI bidding war, and moreover, how to influence it to achieve certain desirable outcomes (e.g. AI for public good and participant compliance). To bridge this gap, this paper proposes a research agenda to develop theoretical models that capture key factors of the AI race, revealing which strategic behaviours may emerge and hypothetical scenarios therein. Strategies from incentive and agreement modelling are directly applicable to systematically analyse how different types of incentives (namely, positive vs. negative, peer vs. institutional, and their combinations) influence safety-compliant behaviours over time, and how such behaviours should be configured to ensure desired global outcomes, studying at the same time how these mechanisms influence AI development. This agenda will provide actionable policies, showing how they need to be employed and deployed in order to achieve compliance and thereby avoid disasters as well as loosing confidence and trust in AI in general.}, note = {Conference: AAAI/ACM Conference on AI, Ethics, and Society(Honolulu, HI, USA)}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } A race for technological supremacy in AI could lead to serious negative consequences, especially whenever ethical and safety procedures are underestimated or even ignored, leading potentially to the rejection of AI in general. For all to enjoy the benefits provided by safe, ethical and trustworthy AI systems, it is crucial to incentivise participants with appropriate strategies that ensure mutually beneficial normative behaviour and safety-compliance from all parties involved. Little attention has been given to understanding the dynamics and emergent behaviours arising from this AI bidding war, and moreover, how to influence it to achieve certain desirable outcomes (e.g. AI for public good and participant compliance). To bridge this gap, this paper proposes a research agenda to develop theoretical models that capture key factors of the AI race, revealing which strategic behaviours may emerge and hypothetical scenarios therein. Strategies from incentive and agreement modelling are directly applicable to systematically analyse how different types of incentives (namely, positive vs. negative, peer vs. institutional, and their combinations) influence safety-compliant behaviours over time, and how such behaviours should be configured to ensure desired global outcomes, studying at the same time how these mechanisms influence AI development. This agenda will provide actionable policies, showing how they need to be employed and deployed in order to achieve compliance and thereby avoid disasters as well as loosing confidence and trust in AI in general. |
Papadimitriou, Sofia; Gazzo, Andrea; Versbraegen, Nassim; Nachtegael, Charlotte; Aerts, Jan; Moreau, Yves; Dooren, Sonia Van; Nowe, Ann; Smits, Guillaume; Lenaerts, Tom Predicting disease-causing variant combinations Journal Article Proceedings of the National Academy of Sciences of the United States of America, 116 (24), pp. 11878-11887, 2019, (DOI: 10.1073/pnas.1815601116). @article{info:hdl:2013/289724, title = {Predicting disease-causing variant combinations}, author = {Sofia Papadimitriou and Andrea Gazzo and Nassim Versbraegen and Charlotte Nachtegael and Jan Aerts and Yves Moreau and Sonia Van Dooren and Ann Nowe and Guillaume Smits and Tom Lenaerts}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/289724/4/doi_273351.pdf}, year = {2019}, date = {2019-01-01}, journal = {Proceedings of the National Academy of Sciences of the United States of America}, volume = {116}, number = {24}, pages = {11878-11887}, abstract = {Notwithstanding important advances in the context of single-variant pathogenicity identification, novel breakthroughs in discerning the origins of many rare diseases require methods able to identify more complex genetic models. We present here the Variant Combinations Pathogenicity Predictor (VarCoPP), a machine-learning approach that identifies pathogenic variant combinations in gene pairs (called digenic or bilocus variant combinations). We show that the results produced by this method are highly accurate and precise, an efficacy that is endorsed when validating the method on recently published independent disease-causing data. Confidence labels of 95% and 99% are identified, representing the probability of a bilocus combination being a true pathogenic result, providing geneticists with rational markers to evaluate the most relevant pathogenic combinations and limit the search space and time. Finally, the VarCoPP has been designed to act as an interpretable method that can provide explanations on why a bilocus combination is predicted as pathogenic and which biological information is important for that prediction. This work provides an important step toward the genetic understanding of rare diseases, paving the way to clinical knowledge and improved patient care.}, note = {DOI: 10.1073/pnas.1815601116}, keywords = {}, pubstate = {published}, tppubtype = {article} } Notwithstanding important advances in the context of single-variant pathogenicity identification, novel breakthroughs in discerning the origins of many rare diseases require methods able to identify more complex genetic models. We present here the Variant Combinations Pathogenicity Predictor (VarCoPP), a machine-learning approach that identifies pathogenic variant combinations in gene pairs (called digenic or bilocus variant combinations). We show that the results produced by this method are highly accurate and precise, an efficacy that is endorsed when validating the method on recently published independent disease-causing data. Confidence labels of 95% and 99% are identified, representing the probability of a bilocus combination being a true pathogenic result, providing geneticists with rational markers to evaluate the most relevant pathogenic combinations and limit the search space and time. Finally, the VarCoPP has been designed to act as an interpretable method that can provide explanations on why a bilocus combination is predicted as pathogenic and which biological information is important for that prediction. This work provides an important step toward the genetic understanding of rare diseases, paving the way to clinical knowledge and improved patient care. |
Renaux, Alexandre; Papadimitriou, Sofia; Versbraegen, Nassim; Nachtegael, Charlotte; Boutry, Simon; Nowé, Ann; Smits, Guillaume; Lenaerts, Tom ORVAL: a novel platform for the prediction and exploration of disease-causing oligogenic variant combinations. Journal Article Nucleic acids research, 47 (W1), pp. W93-W98, 2019, (DOI: 10.1093/nar/gkz437). @article{info:hdl:2013/289958, title = {ORVAL: a novel platform for the prediction and exploration of disease-causing oligogenic variant combinations.}, author = {Alexandre Renaux and Sofia Papadimitriou and Nassim Versbraegen and Charlotte Nachtegael and Simon Boutry and Ann Nowé and Guillaume Smits and Tom Lenaerts}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/289958/4/doi_273585.pdf}, year = {2019}, date = {2019-01-01}, journal = {Nucleic acids research}, volume = {47}, number = {W1}, pages = {W93-W98}, abstract = {A tremendous amount of DNA sequencing data is being produced around the world with the ambition to capture in more detail the mechanisms underlying human diseases. While numerous bioinformatics tools exist that allow the discovery of causal variants in Mendelian diseases, little to no support is provided to do the same for variant combinations, an essential task for the discovery of the causes of oligogenic diseases. ORVAL (the Oligogenic Resource for Variant AnaLysis), which is presented here, provides an answer to this problem by focusing on generating networks of candidate pathogenic variant combinations in gene pairs, as opposed to isolated variants in unique genes. This online platform integrates innovative machine learning methods for combinatorial variant pathogenicity prediction with visualization techniques, offering several interactive and exploratory tools, such as pathogenic gene and protein interaction networks, a ranking of pathogenic gene pairs, as well as visual mappings of the cellular location and pathway information. ORVAL is the first web-based exploration platform dedicated to identifying networks of candidate pathogenic variant combinations with the sole ambition to help in uncovering oligogenic causes for patients that cannot rely on the classical disease analysis tools. ORVAL is available at https://orval.ibsquare.be.}, note = {DOI: 10.1093/nar/gkz437}, keywords = {}, pubstate = {published}, tppubtype = {article} } A tremendous amount of DNA sequencing data is being produced around the world with the ambition to capture in more detail the mechanisms underlying human diseases. While numerous bioinformatics tools exist that allow the discovery of causal variants in Mendelian diseases, little to no support is provided to do the same for variant combinations, an essential task for the discovery of the causes of oligogenic diseases. ORVAL (the Oligogenic Resource for Variant AnaLysis), which is presented here, provides an answer to this problem by focusing on generating networks of candidate pathogenic variant combinations in gene pairs, as opposed to isolated variants in unique genes. This online platform integrates innovative machine learning methods for combinatorial variant pathogenicity prediction with visualization techniques, offering several interactive and exploratory tools, such as pathogenic gene and protein interaction networks, a ranking of pathogenic gene pairs, as well as visual mappings of the cellular location and pathway information. ORVAL is the first web-based exploration platform dedicated to identifying networks of candidate pathogenic variant combinations with the sole ambition to help in uncovering oligogenic causes for patients that cannot rely on the classical disease analysis tools. ORVAL is available at https://orval.ibsquare.be. |
Versbraegen, Nassim; Fouché, Aziz; Nachtegael, Charlotte; Papadimitriou, Sofia; Gazzo, Andrea; Smits, Guillaume; Lenaerts, Tom Using game theory and decision decomposition to effectively discern and characterise bi-locus diseases Journal Article Artificial intelligence in medicine, 99 , 2019, (DOI: 10.1016/j.artmed.2019.06.006). @article{info:hdl:2013/292462, title = {Using game theory and decision decomposition to effectively discern and characterise bi-locus diseases}, author = {Nassim Versbraegen and Aziz Fouché and Charlotte Nachtegael and Sofia Papadimitriou and Andrea Gazzo and Guillaume Smits and Tom Lenaerts}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/292462/5/Elsevier_276089.pdf}, year = {2019}, date = {2019-01-01}, journal = {Artificial intelligence in medicine}, volume = {99}, abstract = {In order to gain insight into oligogenic disorders, understanding those involving bi-locus variant combinations appears to be key. In prior work, we showed that features at multiple biological scales can already be used to discriminate among two types, i.e. disorders involving true digenic and modifier combinations. The current study expands this machine learning work towards dual molecular diagnosis cases, providing a classifier able to effectively distinguish between these three types. To reach this goal and gain an in-depth understanding of the decision process, game theory and tree decomposition techniques are applied to random forest predictors to investigate the relevance of feature combinations in the prediction. A machine learning model with high discrimination capabilities was developed, effectively differentiating the three classes in a biologically meaningful manner. Combining prediction interpretation and statistical analysis, we propose a biologically meaningful characterization of each class relying on specific feature strengths. Figuring out how biological characteristics shift samples towards one of three classes provides clinically relevant insight into the underlying biological processes as well as the disease itself.}, note = {DOI: 10.1016/j.artmed.2019.06.006}, keywords = {}, pubstate = {published}, tppubtype = {article} } In order to gain insight into oligogenic disorders, understanding those involving bi-locus variant combinations appears to be key. In prior work, we showed that features at multiple biological scales can already be used to discriminate among two types, i.e. disorders involving true digenic and modifier combinations. The current study expands this machine learning work towards dual molecular diagnosis cases, providing a classifier able to effectively distinguish between these three types. To reach this goal and gain an in-depth understanding of the decision process, game theory and tree decomposition techniques are applied to random forest predictors to investigate the relevance of feature combinations in the prediction. A machine learning model with high discrimination capabilities was developed, effectively differentiating the three classes in a biologically meaningful manner. Combining prediction interpretation and statistical analysis, we propose a biologically meaningful characterization of each class relying on specific feature strengths. Figuring out how biological characteristics shift samples towards one of three classes provides clinically relevant insight into the underlying biological processes as well as the disease itself. |
Libin, Pieter; Versbraegen, Nassim; Abecasis, Ana A B; Gomes, Perpétua; Lenaerts, Tom; Nowe, Ann Towards a phylogenetic measure to quantify HIV incidence Journal Article CEUR Workshop Proceedings, 2491 , 2019, (Language of publication: en). @article{info:hdl:2013/296964, title = {Towards a phylogenetic measure to quantify HIV incidence}, author = {Pieter Libin and Nassim Versbraegen and Ana A B Abecasis and Perpétua Gomes and Tom Lenaerts and Ann Nowe}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/296964/3/tapm.pdf}, year = {2019}, date = {2019-01-01}, journal = {CEUR Workshop Proceedings}, volume = {2491}, note = {Language of publication: en}, keywords = {}, pubstate = {published}, tppubtype = {article} } |
Saykali, Bechara; Nahaboo, Wallis; Mathiah, Navrita; Racu, Marie-Lucie; Defrance, Matthieu; Migeotte, Isabelle Distinct mesoderm migration phenotypes in extra-embryonic and embryonic regions of the early mouse embryo Journal Article eLife, 8 , 2019, (DOI: 10.7554/eLife.42434.001). @article{info:hdl:2013/282177, title = {Distinct mesoderm migration phenotypes in extra-embryonic and embryonic regions of the early mouse embryo}, author = {Bechara Saykali and Wallis Nahaboo and Navrita Mathiah and Marie-Lucie Racu and Matthieu Defrance and Isabelle Migeotte}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/282177/3/elife-42434-v1.pdf}, year = {2019}, date = {2019-01-01}, journal = {eLife}, volume = {8}, abstract = {In mouse embryo gastrulation, epiblast cells delaminate at the primitive streak to form mesoderm and definitive endoderm, through an epithelial-mesenchymal transition. Mosaic expression of a membrane reporter in nascent mesoderm enabled recording cell shape and trajectory through live imaging. Upon leaving the streak, cells changed shape and extended protrusions of distinct size and abundance depending on the neighboring germ layer, as well as the region of the embryo. Embryonic trajectories were meandrous but directional, while extra-embryonic mesoderm cells showed little net displacement. Embryonic and extra-embryonic mesoderm transcriptomes highlighted distinct guidance, cytoskeleton, adhesion, and extracellular matrix signatures. Specifically, intermediate filaments were highly expressed in extra-embryonic mesoderm, while live imaging for F-actin showed abundance of actin filaments in embryonic mesoderm only. Accordingly, Rhoa or Rac1 conditional deletion in mesoderm inhibited embryonic, but not extra-embryonic mesoderm migration. Overall, this indicates separate cytoskeleton regulation coordinating the morphology and migration of mesoderm subpopulations.}, note = {DOI: 10.7554/eLife.42434.001}, keywords = {}, pubstate = {published}, tppubtype = {article} } In mouse embryo gastrulation, epiblast cells delaminate at the primitive streak to form mesoderm and definitive endoderm, through an epithelial-mesenchymal transition. Mosaic expression of a membrane reporter in nascent mesoderm enabled recording cell shape and trajectory through live imaging. Upon leaving the streak, cells changed shape and extended protrusions of distinct size and abundance depending on the neighboring germ layer, as well as the region of the embryo. Embryonic trajectories were meandrous but directional, while extra-embryonic mesoderm cells showed little net displacement. Embryonic and extra-embryonic mesoderm transcriptomes highlighted distinct guidance, cytoskeleton, adhesion, and extracellular matrix signatures. Specifically, intermediate filaments were highly expressed in extra-embryonic mesoderm, while live imaging for F-actin showed abundance of actin filaments in embryonic mesoderm only. Accordingly, Rhoa or Rac1 conditional deletion in mesoderm inhibited embryonic, but not extra-embryonic mesoderm migration. Overall, this indicates separate cytoskeleton regulation coordinating the morphology and migration of mesoderm subpopulations. |
Istaces, Nicolas; Splittgerber, Marion; Silva, Viviana Lima; Nguyen, Muriel; Thomas, Séverine; Le, Aurore; Achouri, Younes; Calonne, Emilie; Defrance, Matthieu; c c, Fran; Goriely, Stanislas; Azouz, Abdulkader EOMES interacts with RUNX3 and BRG1 to promote innate memory cell formation through epigenetic reprogramming Journal Article Nature communications, 10 (1), 2019, (DOI: 10.1038/s41467-019-11233-6). @article{info:hdl:2013/294385, title = {EOMES interacts with RUNX3 and BRG1 to promote innate memory cell formation through epigenetic reprogramming}, author = {Nicolas Istaces and Marion Splittgerber and Viviana Lima Silva and Muriel Nguyen and Séverine Thomas and Aurore Le and Younes Achouri and Emilie Calonne and Matthieu Defrance and Fran{c c}ois Fuks and Stanislas Goriely and Abdulkader Azouz}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/294385/1/doi_278012.pdf}, year = {2019}, date = {2019-01-01}, journal = {Nature communications}, volume = {10}, number = {1}, abstract = {Memory CD8+ T cells have the ability to provide lifelong immunity against pathogens. Although memory features generally arise after challenge with a foreign antigen, na"ive CD8 single positive (SP) thymocytes may acquire phenotypic and functional characteristics of memory cells in response to cytokines such as interleukin-4. This process is associated with the induction of the T-box transcription factor Eomesodermin (EOMES). However, the underlying molecular mechanisms remain ill-defined. Using epigenomic profiling, we show that these innate memory CD8SP cells acquire only a portion of the active enhancer repertoire of conventional memory cells. This reprograming is secondary to EOMES recruitment, mostly to RUNX3-bound enhancers. Furthermore, EOMES is found within chromatin-associated complexes containing BRG1 and promotes the recruitment of this chromatin remodelling factor. Also, the in vivo acquisition of EOMES-dependent program is BRG1-dependent. In conclusion, our results support a strong epigenetic basis for the EOMES-driven establishment of CD8+ T cell innate memory program.}, note = {DOI: 10.1038/s41467-019-11233-6}, keywords = {}, pubstate = {published}, tppubtype = {article} } Memory CD8+ T cells have the ability to provide lifelong immunity against pathogens. Although memory features generally arise after challenge with a foreign antigen, na"ive CD8 single positive (SP) thymocytes may acquire phenotypic and functional characteristics of memory cells in response to cytokines such as interleukin-4. This process is associated with the induction of the T-box transcription factor Eomesodermin (EOMES). However, the underlying molecular mechanisms remain ill-defined. Using epigenomic profiling, we show that these innate memory CD8SP cells acquire only a portion of the active enhancer repertoire of conventional memory cells. This reprograming is secondary to EOMES recruitment, mostly to RUNX3-bound enhancers. Furthermore, EOMES is found within chromatin-associated complexes containing BRG1 and promotes the recruitment of this chromatin remodelling factor. Also, the in vivo acquisition of EOMES-dependent program is BRG1-dependent. In conclusion, our results support a strong epigenetic basis for the EOMES-driven establishment of CD8+ T cell innate memory program. |
2018 |
Silva, Tiago Chedraoui; Colaprico, Antonio; Olsen, Catharina; Malta, Tathiane Maistro Aistro T M; Bontempi, Gianluca; Ceccarelli, Michele; Berman, Benjamin B P; Noushmehr, Houtan F1000Research, 7 , 2018, (DOI: 10.12688/F1000RESEARCH.14197.1). @article{info:hdl:2013/311579, title = {TCGAbiolinksGUI: A graphical user interface to analyze cancer molecular and clinical data [version 1; peer review: 1 approved, 1 approved with reservations]}, author = {Tiago Chedraoui Silva and Antonio Colaprico and Catharina Olsen and Tathiane Maistro Aistro T M Malta and Gianluca Bontempi and Michele Ceccarelli and Benjamin B P Berman and Houtan Noushmehr}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/311579/1/doi_295223.pdf}, year = {2018}, date = {2018-01-01}, journal = {F1000Research}, volume = {7}, abstract = {The GDC (Genomic Data Commons) data portal provides users with data from cancer genomics studies. Recently, we developed the R/Bioconductor TCGAbiolinks package, which allows users to search, download and prepare cancer genomics data for integrative data analysis. The use of this package requires users to have advanced knowledge of R thus limiting the number of users. To overcome this obstacle and improve the accessibility of the package by a wider range of users, we developed a graphical user interface (GUI) using Shiny available through the package TCGAbiolinksGUI. The TCGAbiolinksGUI package is freely available within the Bioconductor project at http://bioconductor.org/packages/TCGAbiolinksGUI/. Links to the GitHub repository, a demo version of the tool, a docker image and PDF/video tutorials are available from the TCGAbiolinksGUI site.}, note = {DOI: 10.12688/F1000RESEARCH.14197.1}, keywords = {}, pubstate = {published}, tppubtype = {article} } The GDC (Genomic Data Commons) data portal provides users with data from cancer genomics studies. Recently, we developed the R/Bioconductor TCGAbiolinks package, which allows users to search, download and prepare cancer genomics data for integrative data analysis. The use of this package requires users to have advanced knowledge of R thus limiting the number of users. To overcome this obstacle and improve the accessibility of the package by a wider range of users, we developed a graphical user interface (GUI) using Shiny available through the package TCGAbiolinksGUI. The TCGAbiolinksGUI package is freely available within the Bioconductor project at http://bioconductor.org/packages/TCGAbiolinksGUI/. Links to the GitHub repository, a demo version of the tool, a docker image and PDF/video tutorials are available from the TCGAbiolinksGUI site. |
Cava, Claudia; Bertoli, Gloria; Colaprico, Antonio; Olsen, Catharina; Bontempi, Gianluca; Castiglioni, Isabella Integration of multiple networks and pathways identifies cancer driver genes in pan-cancer analysis Journal Article BMC genomics, 19 (1), 2018, (DOI: 10.1186/s12864-017-4423-x). @article{info:hdl:2013/268430, title = {Integration of multiple networks and pathways identifies cancer driver genes in pan-cancer analysis}, author = {Claudia Cava and Gloria Bertoli and Antonio Colaprico and Catharina Olsen and Gianluca Bontempi and Isabella Castiglioni}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/268430/1/doi_252057.pdf}, year = {2018}, date = {2018-01-01}, journal = {BMC genomics}, volume = {19}, number = {1}, abstract = {Background: Modern high-throughput genomic technologies represent a comprehensive hallmark of molecular changes in pan-cancer studies. Although different cancer gene signatures have been revealed, the mechanism of tumourigenesis has yet to be completely understood. Pathways and networks are important tools to explain the role of genes in functional genomic studies. However, few methods consider the functional non-equal roles of genes in pathways and the complex gene-gene interactions in a network. Results: We present a novel method in pan-cancer analysis that identifies de-regulated genes with a functional role by integrating pathway and network data. A pan-cancer analysis of 7158 tumour/normal samples from 16 cancer types identified 895 genes with a central role in pathways and de-regulated in cancer. Comparing our approach with 15 current tools that identify cancer driver genes, we found that 35.6% of the 895 genes identified by our method have been found as cancer driver genes with at least 2/15 tools. Finally, we applied a machine learning algorithm on 16 independent GEO cancer datasets to validate the diagnostic role of cancer driver genes for each cancer. We obtained a list of the top-ten cancer driver genes for each cancer considered in this study. Conclusions: Our analysis 1) confirmed that there are several known cancer driver genes in common among different types of cancer, 2) highlighted that cancer driver genes are able to regulate crucial pathways.}, note = {DOI: 10.1186/s12864-017-4423-x}, keywords = {}, pubstate = {published}, tppubtype = {article} } Background: Modern high-throughput genomic technologies represent a comprehensive hallmark of molecular changes in pan-cancer studies. Although different cancer gene signatures have been revealed, the mechanism of tumourigenesis has yet to be completely understood. Pathways and networks are important tools to explain the role of genes in functional genomic studies. However, few methods consider the functional non-equal roles of genes in pathways and the complex gene-gene interactions in a network. Results: We present a novel method in pan-cancer analysis that identifies de-regulated genes with a functional role by integrating pathway and network data. A pan-cancer analysis of 7158 tumour/normal samples from 16 cancer types identified 895 genes with a central role in pathways and de-regulated in cancer. Comparing our approach with 15 current tools that identify cancer driver genes, we found that 35.6% of the 895 genes identified by our method have been found as cancer driver genes with at least 2/15 tools. Finally, we applied a machine learning algorithm on 16 independent GEO cancer datasets to validate the diagnostic role of cancer driver genes for each cancer. We obtained a list of the top-ten cancer driver genes for each cancer considered in this study. Conclusions: Our analysis 1) confirmed that there are several known cancer driver genes in common among different types of cancer, 2) highlighted that cancer driver genes are able to regulate crucial pathways. |
Reggiani, Claudio; "e, Yann-A; Bontempi, Gianluca Feature selection in high-dimensional dataset using MapReduce Journal Article Communications in computer and information science, 823 , pp. 101-115, 2018, (DOI: 10.1007/978-3-319-76892-2_8). @article{info:hdl:2013/269455, title = {Feature selection in high-dimensional dataset using MapReduce}, author = {Claudio Reggiani and Yann-A{"e}l Le Borgne and Gianluca Bontempi}, url = {http://hdl.handle.net/2013/ULB-DIPOT:oai:dipot.ulb.ac.be:2013/269455}, year = {2018}, date = {2018-01-01}, journal = {Communications in computer and information science}, volume = {823}, pages = {101-115}, abstract = {This paper describes a distributed MapReduce implementation of the minimum Redundancy Maximum Relevance algorithm, a popular feature selection method in bioinformatics and network inference problems. The proposed approach handles both tall/narrow and wide/short datasets. We further provide an open source implementation based on Hadoop/Spark, and illustrate its scalability on datasets involving millions of observations or features.}, note = {DOI: 10.1007/978-3-319-76892-2_8}, keywords = {}, pubstate = {published}, tppubtype = {article} } This paper describes a distributed MapReduce implementation of the minimum Redundancy Maximum Relevance algorithm, a popular feature selection method in bioinformatics and network inference problems. The proposed approach handles both tall/narrow and wide/short datasets. We further provide an open source implementation based on Hadoop/Spark, and illustrate its scalability on datasets involving millions of observations or features. |
Trepo, Eric; Goossens, Nicolas; Fujiwara, Naoto; Song, Won-Min; Colaprico, Antonio; Marot, Astrid; Spahr, Laurent; Demetter, Pieter; Sempoux, Christine; Im, Gene Y; Saldarriaga, Joan; Gustot, Thierry; `e, Jacques Devi; Thung, Swan SN; Minsart, Charlotte; Serste, Thomas; Bontempi, Gianluca; Abdelrahman, Karim; Henrion, Jean; Degré, Delphine; Lucidi, Valerio; Rubbia-Brandt, Laura; Nair, Venugopalan D; Moreno, Christophe; Deltenre, Pierre; Hoshida, Yujin; Franchimont, Denis Gastroenterology, 154 (4), pp. 965-975, 2018, (DOI: 10.1053/j.gastro.2017.10.048). @article{info:hdl:2013/269084, title = {Combination of Gene Expression Signature and Model for End-Stage Liver Disease Score Predicts Survival of Patients With Severe Alcoholic Hepatitis}, author = {Eric Trepo and Nicolas Goossens and Naoto Fujiwara and Won-Min Song and Antonio Colaprico and Astrid Marot and Laurent Spahr and Pieter Demetter and Christine Sempoux and Gene Y Im and Joan Saldarriaga and Thierry Gustot and Jacques Devi{`e}re and Swan SN Thung and Charlotte Minsart and Thomas Serste and Gianluca Bontempi and Karim Abdelrahman and Jean Henrion and Delphine Degré and Valerio Lucidi and Laura Rubbia-Brandt and Venugopalan D Nair and Christophe Moreno and Pierre Deltenre and Yujin Hoshida and Denis Franchimont}, url = {http://hdl.handle.net/2013/ULB-DIPOT:oai:dipot.ulb.ac.be:2013/269084}, year = {2018}, date = {2018-01-01}, journal = {Gastroenterology}, volume = {154}, number = {4}, pages = {965-975}, abstract = {Background & Aims: Patients with severe alcoholic hepatitis (AH) have a high risk of death within 90 days. Corticosteroids, which can cause severe adverse events, are the only treatment that increases short-term survival. It is a challenge to predict outcomes of patients with severe AH. Therefore, we developed a scoring system to predict patient survival, integrating baseline molecular and clinical variables. Methods: We obtained fixed liver biopsy samples from 71 consecutive patients diagnosed with severe AH and treated with corticosteroids from July 2006 through December 2013 in Brussels, Belgium (derivation cohort). Gene expression patterns were analyzed by microarrays and clinical data were collected for 180 days. We identified gene expression signatures and clinical data that are associated with survival without liver transplantation at 90 and 180 days after initiation of corticosteroid therapy. Findings were validated using liver biopsies from 48 consecutive patients with severe AH treated with corticosteroids, collected from March 2010 through February 2015 at hospitals in Belgium and Switzerland (validation cohort 1) and in liver biopsies from 20 patients (9 received corticosteroid treatment), collected from January 2012 through May 2015 in the United States (validation cohort 2). Results: We integrated data on expression patterns of 123 genes and the model for end-stage liver disease (MELD) scores to assign patients to groups with poor survival (29% survived 90 days and 26% survived 180 days) and good survival (76% survived 90 days and 65% survived 180 days) (P <.001) in the derivation cohort. We named this assignment system the gene signature--MELD (gs-MELD) score. In validation cohort 1, the gs-MELD score discriminated patients with poor survival (43% survived 90 days) from those with good survival (96% survived 90 days) (P <.001). The gs-MELD score also discriminated between patients with a poor survival at 180 days (34% survived) and a good survival at 180 days (84% survived) (P <.001). The time-dependent area under the receiver operator characteristic curve for the score was 0.86 (95% confidence interval 0.73--0.99) for survival at 90 days, and 0.83 (95% confidence interval 0.71--0.96) for survival at 180 days. This score outperformed other clinical models to predict survival of patients with severe AH in validation cohort 1. In validation cohort 2, the gs-MELD discriminated patients with a poor survival at 90 days (12% survived) from those with a good survival at 90 days (100%) (P <.001). Conclusions: We integrated data on baseline liver gene expression pattern and the MELD score to create the gs-MELD scoring system, which identifies patients with severe AH, treated or not with corticosteroids, most and least likely to survive for 90 and 180 days.}, note = {DOI: 10.1053/j.gastro.2017.10.048}, keywords = {}, pubstate = {published}, tppubtype = {article} } Background & Aims: Patients with severe alcoholic hepatitis (AH) have a high risk of death within 90 days. Corticosteroids, which can cause severe adverse events, are the only treatment that increases short-term survival. It is a challenge to predict outcomes of patients with severe AH. Therefore, we developed a scoring system to predict patient survival, integrating baseline molecular and clinical variables. Methods: We obtained fixed liver biopsy samples from 71 consecutive patients diagnosed with severe AH and treated with corticosteroids from July 2006 through December 2013 in Brussels, Belgium (derivation cohort). Gene expression patterns were analyzed by microarrays and clinical data were collected for 180 days. We identified gene expression signatures and clinical data that are associated with survival without liver transplantation at 90 and 180 days after initiation of corticosteroid therapy. Findings were validated using liver biopsies from 48 consecutive patients with severe AH treated with corticosteroids, collected from March 2010 through February 2015 at hospitals in Belgium and Switzerland (validation cohort 1) and in liver biopsies from 20 patients (9 received corticosteroid treatment), collected from January 2012 through May 2015 in the United States (validation cohort 2). Results: We integrated data on expression patterns of 123 genes and the model for end-stage liver disease (MELD) scores to assign patients to groups with poor survival (29% survived 90 days and 26% survived 180 days) and good survival (76% survived 90 days and 65% survived 180 days) (P <.001) in the derivation cohort. We named this assignment system the gene signature--MELD (gs-MELD) score. In validation cohort 1, the gs-MELD score discriminated patients with poor survival (43% survived 90 days) from those with good survival (96% survived 90 days) (P <.001). The gs-MELD score also discriminated between patients with a poor survival at 180 days (34% survived) and a good survival at 180 days (84% survived) (P <.001). The time-dependent area under the receiver operator characteristic curve for the score was 0.86 (95% confidence interval 0.73--0.99) for survival at 90 days, and 0.83 (95% confidence interval 0.71--0.96) for survival at 180 days. This score outperformed other clinical models to predict survival of patients with severe AH in validation cohort 1. In validation cohort 2, the gs-MELD discriminated patients with a poor survival at 90 days (12% survived) from those with a good survival at 90 days (100%) (P <.001). Conclusions: We integrated data on baseline liver gene expression pattern and the MELD score to create the gs-MELD scoring system, which identifies patients with severe AH, treated or not with corticosteroids, most and least likely to survive for 90 and 180 days. |
Cava, Claudia; Bertoli, Gloria; Colaprico, Antonio; Bontempi, Gianluca; Mauri, Giancarlo; Castiglioni, Isabella In-silico integration approach to identify a key miRNA regulating a gene network in aggressive prostate cancer Journal Article International journal of molecular sciences, 19 (3), 2018, (DOI: 10.3390/ijms19030910). @article{info:hdl:2013/270452, title = {In-silico integration approach to identify a key miRNA regulating a gene network in aggressive prostate cancer}, author = {Claudia Cava and Gloria Bertoli and Antonio Colaprico and Gianluca Bontempi and Giancarlo Mauri and Isabella Castiglioni}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/270452/1/doi_254079.pdf}, year = {2018}, date = {2018-01-01}, journal = {International journal of molecular sciences}, volume = {19}, number = {3}, abstract = {Like other cancer diseases, prostate cancer (PC) is caused by the accumulation of genetic alterations in the cells that drives malignant growth. These alterations are revealed by gene profiling and copy number alteration (CNA) analysis. Moreover, recent evidence suggests that also microRNAs have an important role in PC development. Despite efforts to profile PC, the alterations (gene, CNA, and miRNA) and biological processes that correlate with disease development and progression remain partially elusive. Many gene signatures proposed as diagnostic or prognostic tools in cancer poorly overlap. The identification of co-expressed genes, that are functionally related, can identify a core network of genes associated with PC with a better reproducibility. By combining different approaches, including the integration of mRNA expression profiles, CNAs, and miRNA expression levels, we identified a gene signature of four genes overlapping with other published gene signatures and able to distinguish, in silico, high Gleason-scored PC from normal human tissue, which was further enriched to 19 genes by gene co-expression analysis. From the analysis of miRNAs possibly regulating this network, we found that hsa-miR-153 was highly connected to the genes in the network. Our results identify a four-gene signature with diagnostic and prognostic value in PC and suggest an interesting gene network that could play a key regulatory role in PC development and progression. Furthermore, hsa-miR-153, controlling this network, could be a potential biomarker for theranostics in high Gleason-scored PC.}, note = {DOI: 10.3390/ijms19030910}, keywords = {}, pubstate = {published}, tppubtype = {article} } Like other cancer diseases, prostate cancer (PC) is caused by the accumulation of genetic alterations in the cells that drives malignant growth. These alterations are revealed by gene profiling and copy number alteration (CNA) analysis. Moreover, recent evidence suggests that also microRNAs have an important role in PC development. Despite efforts to profile PC, the alterations (gene, CNA, and miRNA) and biological processes that correlate with disease development and progression remain partially elusive. Many gene signatures proposed as diagnostic or prognostic tools in cancer poorly overlap. The identification of co-expressed genes, that are functionally related, can identify a core network of genes associated with PC with a better reproducibility. By combining different approaches, including the integration of mRNA expression profiles, CNAs, and miRNA expression levels, we identified a gene signature of four genes overlapping with other published gene signatures and able to distinguish, in silico, high Gleason-scored PC from normal human tissue, which was further enriched to 19 genes by gene co-expression analysis. From the analysis of miRNAs possibly regulating this network, we found that hsa-miR-153 was highly connected to the genes in the network. Our results identify a four-gene signature with diagnostic and prognostic value in PC and suggest an interesting gene network that could play a key regulatory role in PC development and progression. Furthermore, hsa-miR-153, controlling this network, could be a potential biomarker for theranostics in high Gleason-scored PC. |
Ioannidis, J P A; Bhattacharya, S; Evers, J L H; Veen, Der F V; Somigliana, E; Barratt, C L R; Bontempi, Gianluca; Baird, D T; Crosignani, P; Devroey, P; Diedrich, Klaus; Farquharson, R G; Fraser, L R; Geraedts, Joep Pm M; Gianaroli, Luca; Vecchia, La C; Magli, C; Negri, E; Sunde, A; Tapanainen, J S; Tarlatzis, Basil; Steirteghem, A V; Veiga, A Protect us from poor-quality medical research Journal Article Human reproduction, 33 (5), pp. 770-776, 2018, (DOI: 10.1093/humrep/dey056). @article{info:hdl:2013/272828, title = {Protect us from poor-quality medical research}, author = {J P A Ioannidis and S Bhattacharya and J L H Evers and F V Der Veen and E Somigliana and C L R Barratt and Gianluca Bontempi and D T Baird and P Crosignani and P Devroey and Klaus Diedrich and R G Farquharson and L R Fraser and Joep Pm M Geraedts and Luca Gianaroli and C La Vecchia and C Magli and E Negri and A Sunde and J S Tapanainen and Basil Tarlatzis and A V Steirteghem and A Veiga}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/272828/1/doi_256455.pdf}, year = {2018}, date = {2018-01-01}, journal = {Human reproduction}, volume = {33}, number = {5}, pages = {770-776}, abstract = {Much of the published medical research is apparently flawed, cannot be replicated and/or has limited or no utility. This article presents an overview of the current landscape of biomedical research, identifies problems associated with common study designs and considers potential solutions. Randomized clinical trials, observational studies, systematic reviews and meta-analyses are discussed in terms of their inherent limitations and potential ways of improving their conduct, analysis and reporting. The current emphasis on statistical significance needs to be replaced by sound design, transparency and willingness to share data with a clear commitment towards improving the quality and utility of clinical research.}, note = {DOI: 10.1093/humrep/dey056}, keywords = {}, pubstate = {published}, tppubtype = {article} } Much of the published medical research is apparently flawed, cannot be replicated and/or has limited or no utility. This article presents an overview of the current landscape of biomedical research, identifies problems associated with common study designs and considers potential solutions. Randomized clinical trials, observational studies, systematic reviews and meta-analyses are discussed in terms of their inherent limitations and potential ways of improving their conduct, analysis and reporting. The current emphasis on statistical significance needs to be replaced by sound design, transparency and willingness to share data with a clear commitment towards improving the quality and utility of clinical research. |
Carcillo, Fabrizio; "e, Yann-A; Caelen, Olivier; Bontempi, Gianluca Correction to: Streaming active learning strategies for real-life credit detection: assessment and visualization Journal Article International journal of data science and analytics (Print), 5 (4), pp. 301-302, 2018, (DOI: 10.1007/s41060-018-0123-0). @article{info:hdl:2013/311389, title = {Correction to: Streaming active learning strategies for real-life credit detection: assessment and visualization}, author = {Fabrizio Carcillo and Yann-A{"e}l Le Borgne and Olivier Caelen and Gianluca Bontempi}, url = {http://hdl.handle.net/2013/ULB-DIPOT:oai:dipot.ulb.ac.be:2013/311389}, year = {2018}, date = {2018-01-01}, journal = {International journal of data science and analytics (Print)}, volume = {5}, number = {4}, pages = {301-302}, abstract = {A line was misplaced in Table 1. The correct Table 1 is given below. The original article has been corrected. (Table presented.).}, note = {DOI: 10.1007/s41060-018-0123-0}, keywords = {}, pubstate = {published}, tppubtype = {article} } A line was misplaced in Table 1. The correct Table 1 is given below. The original article has been corrected. (Table presented.). |
Carcillo, Fabrizio; "e, Yann-A; Caelen, Olivier; Bontempi, Gianluca Streaming active learning strategies for real-life credit card fraud detection: assessment and visualization Journal Article International journal of data science and analytics (Print), 5 (4), pp. 285-300, 2018, (DOI: 10.1007/s41060-018-0116-z). @article{info:hdl:2013/311546, title = {Streaming active learning strategies for real-life credit card fraud detection: assessment and visualization}, author = {Fabrizio Carcillo and Yann-A{"e}l Le Borgne and Olivier Caelen and Gianluca Bontempi}, url = {http://hdl.handle.net/2013/ULB-DIPOT:oai:dipot.ulb.ac.be:2013/311546}, year = {2018}, date = {2018-01-01}, journal = {International journal of data science and analytics (Print)}, volume = {5}, number = {4}, pages = {285-300}, abstract = {Credit card fraud detection is a very challenging problem because of the specific nature of transaction data and the labeling process. The transaction data are peculiar because they are obtained in a streaming fashion, and they are strongly imbalanced and prone to non-stationarity. The labeling is the outcome of an active learning process, as every day human investigators contact only a small number of cardholders (associated with the riskiest transactions) and obtain the class (fraud or genuine) of the related transactions. An adequate selection of the set of cardholders is therefore crucial for an efficient fraud detection process. In this paper, we present a number of active learning strategies and we investigate their fraud detection accuracies. We compare different criteria (supervised, semi-supervised and unsupervised) to query unlabeled transactions. Finally, we highlight the existence of an exploitation/exploration trade-off for active learning in the context of fraud detection, which has so far been overlooked in the literature.}, note = {DOI: 10.1007/s41060-018-0116-z}, keywords = {}, pubstate = {published}, tppubtype = {article} } Credit card fraud detection is a very challenging problem because of the specific nature of transaction data and the labeling process. The transaction data are peculiar because they are obtained in a streaming fashion, and they are strongly imbalanced and prone to non-stationarity. The labeling is the outcome of an active learning process, as every day human investigators contact only a small number of cardholders (associated with the riskiest transactions) and obtain the class (fraud or genuine) of the related transactions. An adequate selection of the set of cardholders is therefore crucial for an efficient fraud detection process. In this paper, we present a number of active learning strategies and we investigate their fraud detection accuracies. We compare different criteria (supervised, semi-supervised and unsupervised) to query unlabeled transactions. Finally, we highlight the existence of an exploitation/exploration trade-off for active learning in the context of fraud detection, which has so far been overlooked in the literature. |
de Bony, Eric James; Bizet, Martin; Grembergen, Olivier Van; Hassabi, Bouchra; Calonne, Emilie; Putmans, Pascale; Bontempi, Gianluca; c c, Fran Comprehensive identification of long noncoding RNAs in colorectal cancer Journal Article Oncotarget, 9 (45), pp. 27605-27629, 2018, (DOI: 10.18632/oncotarget.25218). @article{info:hdl:2013/278063, title = {Comprehensive identification of long noncoding RNAs in colorectal cancer}, author = {Eric James de Bony and Martin Bizet and Olivier Van Grembergen and Bouchra Hassabi and Emilie Calonne and Pascale Putmans and Gianluca Bontempi and Fran{c c}ois Fuks}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/278063/3/doi_261690.pdf}, year = {2018}, date = {2018-01-01}, journal = {Oncotarget}, volume = {9}, number = {45}, pages = {27605-27629}, abstract = {Colorectal cancer (CRC) is one of the most common cancers in humans and a leading cause of cancer-related deaths worldwide. As in the case of other cancers, CRC heterogeneity leads to a wide range of clinical outcomes and complicates therapy. Over the years, multiple factors have emerged as markers of CRC heterogeneity, improving tumor classification and selection of therapeutic strategies. Understanding the molecular mechanisms underlying this heterogeneity remains a major challenge. A considerable research effort is therefore devoted to identifying additional features of colorectal tumors, in order to better understand CRC etiology and to multiply therapeutic avenues. Recently, long noncoding RNAs (lncRNAs) have emerged as important players in physiological and pathological processes, including CRC. Here we looked for lncRNAs that might contribute to the various colorectal tumor phenotypes. We thus monitored the expression of 4898 lncRNA genes across 566 CRC samples and identified 282 lncRNAs reflecting CRC heterogeneity. We then inferred potential functions of these lncRNAs. Our results highlight lncRNAs that may participate in the major processes altered in distinct CRC cases, such as WNT/β-catenin and TGF-β signaling, immunity, the epithelial-to-mesenchymal transition (EMT), and angiogenesis. For several candidates, we provide experimental evidence supporting our functional predictions that they may be involved in the cell cycle or the EMT. Overall, our work identifies lncRNAs associated with key CRC characteristics and provides insights into their respective functions. Our findings constitute a further step towards understanding the contribution of lncRNAs to CRC heterogeneity. They may open new therapeutic opportunities.}, note = {DOI: 10.18632/oncotarget.25218}, keywords = {}, pubstate = {published}, tppubtype = {article} } Colorectal cancer (CRC) is one of the most common cancers in humans and a leading cause of cancer-related deaths worldwide. As in the case of other cancers, CRC heterogeneity leads to a wide range of clinical outcomes and complicates therapy. Over the years, multiple factors have emerged as markers of CRC heterogeneity, improving tumor classification and selection of therapeutic strategies. Understanding the molecular mechanisms underlying this heterogeneity remains a major challenge. A considerable research effort is therefore devoted to identifying additional features of colorectal tumors, in order to better understand CRC etiology and to multiply therapeutic avenues. Recently, long noncoding RNAs (lncRNAs) have emerged as important players in physiological and pathological processes, including CRC. Here we looked for lncRNAs that might contribute to the various colorectal tumor phenotypes. We thus monitored the expression of 4898 lncRNA genes across 566 CRC samples and identified 282 lncRNAs reflecting CRC heterogeneity. We then inferred potential functions of these lncRNAs. Our results highlight lncRNAs that may participate in the major processes altered in distinct CRC cases, such as WNT/β-catenin and TGF-β signaling, immunity, the epithelial-to-mesenchymal transition (EMT), and angiogenesis. For several candidates, we provide experimental evidence supporting our functional predictions that they may be involved in the cell cycle or the EMT. Overall, our work identifies lncRNAs associated with key CRC characteristics and provides insights into their respective functions. Our findings constitute a further step towards understanding the contribution of lncRNAs to CRC heterogeneity. They may open new therapeutic opportunities. |
Stefani, Jacopo De; "e, Yann-A; Caelen, Olivier; Hattab, Dalila; Bontempi, Gianluca Batch and incremental dynamic factor machine learning for multivariate and multi-step-ahead forecasting Journal Article International journal of data science and analytics (Print), 7 (4), pp. 311-329, 2018, (DOI: 10.1007/s41060-018-0150-x). @article{info:hdl:2013/283230, title = {Batch and incremental dynamic factor machine learning for multivariate and multi-step-ahead forecasting}, author = {Jacopo De Stefani and Yann-A{"e}l Le Borgne and Olivier Caelen and Dalila Hattab and Gianluca Bontempi}, url = {http://hdl.handle.net/2013/ULB-DIPOT:oai:dipot.ulb.ac.be:2013/283230}, year = {2018}, date = {2018-01-01}, journal = {International journal of data science and analytics (Print)}, volume = {7}, number = {4}, pages = {311-329}, abstract = {Most multivariate forecasting methods in the literature are restricted to vector time series of low dimension, linear methods and short horizons. Big data revolution is instead shifting the focus to problems (e.g., issued from the IoT technology) characterized by very large dimension, nonlinearity and long forecasting horizons. This paper discusses and compares a set of state-of-the-art methods which could be promising in tackling such challenges. Also, it proposes DFML, a machine-learning version of the dynamic factor model, a successful forecasting methodology well-known in econometrics. The DFML strategy is based on a out-of-sample selection of the nonlinear forecaster, the number of latent components and the multi-step-ahead strategy. We will discuss both a batch and an incremental version of DFML, and we will show that it can consistently outperform state-of-the-art methods in a number of Synthetic and real forecasting tasks.}, note = {DOI: 10.1007/s41060-018-0150-x}, keywords = {}, pubstate = {published}, tppubtype = {article} } Most multivariate forecasting methods in the literature are restricted to vector time series of low dimension, linear methods and short horizons. Big data revolution is instead shifting the focus to problems (e.g., issued from the IoT technology) characterized by very large dimension, nonlinearity and long forecasting horizons. This paper discusses and compares a set of state-of-the-art methods which could be promising in tackling such challenges. Also, it proposes DFML, a machine-learning version of the dynamic factor model, a successful forecasting methodology well-known in econometrics. The DFML strategy is based on a out-of-sample selection of the nonlinear forecaster, the number of latent components and the multi-step-ahead strategy. We will discuss both a batch and an incremental version of DFML, and we will show that it can consistently outperform state-of-the-art methods in a number of Synthetic and real forecasting tasks. |
Abels, Axel; Roijers, Diederik D M; Lenaerts, Tom; Nowe, Ann; Steckelmacher, Denis Dynamic Weights in Multi-Objective Deep Reinforcement Learning Inproceedings Proceedings of the 30th Benelux Conference on Artificial Intelligence, pp. 1-2, Springer, 2018, (Language of publication: en). @inproceedings{info:hdl:2013/291980, title = {Dynamic Weights in Multi-Objective Deep Reinforcement Learning}, author = {Axel Abels and Diederik D M Roijers and Tom Lenaerts and Ann Nowe and Denis Steckelmacher}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/291980/3/short.pdf}, year = {2018}, date = {2018-01-01}, booktitle = {Proceedings of the 30th Benelux Conference on Artificial Intelligence}, pages = {1-2}, publisher = {Springer}, series = {CCIS series}, note = {Language of publication: en}, keywords = {}, pubstate = {published}, tppubtype = {inproceedings} } |
`e, Nathaniel Mon P; Lenaerts, Tom; Pacheco, Jorge J M; Dingli, David Evolutionary Dynamics of Paroxysmal Nocturnal Hemoglobinuria Journal Article PLoS computational biology, 14 (6), 2018, (DOI: 10.1371/journal.pcbi.1006133). @article{info:hdl:2013/267360, title = {Evolutionary Dynamics of Paroxysmal Nocturnal Hemoglobinuria}, author = {Nathaniel Mon P{`e}re and Tom Lenaerts and Jorge J M Pacheco and David Dingli}, url = {https://dipot.ulb.ac.be/dspace/bitstream/2013/267360/4/doi_250987.pdf}, year = {2018}, date = {2018-01-01}, journal = {PLoS computational biology}, volume = {14}, number = {6}, abstract = {Paroxysmal nocturnal hemoglobinuria (PNH) is an acquired clonal blood disorder characterized by hemolysis and a high risk of thrombosis, that is due to a deficiency in several cell surface proteins that prevent complement activation. Its origin has been traced to a somatic mutation in the PIG-A gene within hematopoietic stem cells (HSC). However, to date the question of how this mutant clone expands in size to contribute significantly to hematopoiesis remains under debate. One hypothesis posits the existence of a selective advantage of PIG-A mutated cells due to an immune mediated attack on normal HSC, but the evidence supporting this hypothesis is inconclusive. An alternative (and simpler) explanation attributes clonal expansion to neutral drift, in which case selection neither favours nor inhibits expansion of PIG-A mutated HSC. Here we examine the implications of the neutral drift model by numerically evolving a Markov chain for the probabilities of all possible outcomes, and investigate the possible occurrence and evolution, within this framework, of multiple independently arising clones within the HSC pool. Predictions of the model agree well with the known incidence of the disease and average age at diagnosis. Notwithstanding the slight difference in clonal expansion rates between our results and those reported in the literature, our model results lead to a relative stability of clone size when averaging multiple cases, in accord with what has been observed in human trials. The probability of a patient harbouring a second clone in the HSC pool was found to be extremely low (~10-8). Thus our results suggest that in clinical cases of PNH where two independent clones of mutant cells are observed, only one of those is likely to have originated in the HSC pool.}, note = {DOI: 10.1371/journal.pcbi.1006133}, keywords = {}, pubstate = {published}, tppubtype = {article} } Paroxysmal nocturnal hemoglobinuria (PNH) is an acquired clonal blood disorder characterized by hemolysis and a high risk of thrombosis, that is due to a deficiency in several cell surface proteins that prevent complement activation. Its origin has been traced to a somatic mutation in the PIG-A gene within hematopoietic stem cells (HSC). However, to date the question of how this mutant clone expands in size to contribute significantly to hematopoiesis remains under debate. One hypothesis posits the existence of a selective advantage of PIG-A mutated cells due to an immune mediated attack on normal HSC, but the evidence supporting this hypothesis is inconclusive. An alternative (and simpler) explanation attributes clonal expansion to neutral drift, in which case selection neither favours nor inhibits expansion of PIG-A mutated HSC. Here we examine the implications of the neutral drift model by numerically evolving a Markov chain for the probabilities of all possible outcomes, and investigate the possible occurrence and evolution, within this framework, of multiple independently arising clones within the HSC pool. Predictions of the model agree well with the known incidence of the disease and average age at diagnosis. Notwithstanding the slight difference in clonal expansion rates between our results and those reported in the literature, our model results lead to a relative stability of clone size when averaging multiple cases, in accord with what has been observed in human trials. The probability of a patient harbouring a second clone in the HSC pool was found to be extremely low (~10-8). Thus our results suggest that in clinical cases of PNH where two independent clones of mutant cells are observed, only one of those is likely to have originated in the HSC pool. |